The Year of Everything
Everything that happens to people happens on a schedule. This is the schedule.
Published by Dustin Cole Data
- Births counted
- 85,712,738
- Deaths counted
- 19,433,596
- Years of record
- 1871–2019
Everything that happens to people happens on a schedule. This is the schedule.
Published by Dustin Cole Data
Forty percent fewer American babies arrive on Christmas Day than on an ordinary day. It is the emptiest date of the whole year — and three days later there are eight percent too many.
Point at any date, or pick one from the list below, to read it off the record.
Every square is one date of the year. Its colour is how many babies were born on that date, next to an ordinary day. The two ringed squares are the emptiest date of the year (December 25) and the fullest (September 9).
10% or more below · 12 days5 to 10% below · 8 days2 to 5% below · 79 daysa little below · 95 daysa little above · 61 days2 to 5% above · 55 days5% or more above · 56 days85,712,738 American births, 1994–2014, from CDC/NCHS and SSA daily counts via the FiveThirtyEight births archive. Each date is compared with an average day of the same year and the same weekday, so no date is helped or hurt by falling on a Sunday more often.
Every date the year has a name for, and how far it sits from an ordinary day. Under each month is that whole month’s own average.
The month overall −3.8%
The month overall −1.7%
The month overall −1.5%
No date of its own.
The month overall −2.6%
The month overall −1.5%
The month overall +1.4%
No date of its own.
The month overall +3.7%
The month overall +5.0%
No date of its own.
The month overall +5.8%
No date of its own.
The month overall +0.4%
The month overall −2.3%
The month overall −1.6%
The calendar only moves on days we have a name for. Christmas, New Year’s, the Fourth of July, Thanksgiving, Halloween. Then the babies come back: the days straight after each one run above normal.
September is the fullest stretch of the year — every day from the 9th to the 24th sits eight to ten percent above an ordinary day. And the 13th of the month is the quietest date of any month, about three percent below the 12th and the 14th.
Of the thirty-one days of the month, the thirteenth is the least popular day to get married and the least common day to be born. It is the same superstition measured twice — once on a date people pick, and once on a date they mostly cannot.
WeddingsConnecticut, 1900–2001 — a date two people chose
The thirteenth: 67.1% of an ordinary day
BirthsUnited States, 1994–2014 — a date almost nobody chose
The thirteenth: 96.5% of an ordinary day
Both combs are drawn to one scale: 60 to 130, where 100 is a matched ordinary day of the same month, year and weekday.
In 102 years of Connecticut weddings, 58,960 took place on a thirteenth where 87,841 ordinary days would have brought — 28,881 weddings that went somewhere else on the calendar. It holds in every era, and on every day of the week: the thirteenth is avoided whether it falls on a Friday or not.
Babies cannot read a calendar, and it shows. The thirteenth is still the quietest day of the month for American births — but by 3.5%, not 32.9%.
Births on the thirteenth, split by the weekday it fell on. A different scale from the combs above: this whole chart would fit inside one of their teeth.
Bars hang from 100 — an ordinary day of the same weekday and month — down to a floor of 92. A Friday the thirteenth brings 6.4% fewer American babies than an ordinary Friday of the same month. A Saturday the thirteenth brings 1.1% fewer than an ordinary Saturday. Weekdays are when a birth is most likely to have been booked in advance.
New York City, 1871 to 1945: 2,904,802 death certificates, one band a year. The year turns over — summer empties out, winter fills up — and the reason was the babies.
Point at a band, or tap one, to read that year off the record.
New York City Municipal Archives, Index to Digitized Death Certificates, via NYC Open Data.
100 is an average day of that same year.
| Years | JulyJul | JanuaryJan | 70100150 |
|---|---|---|---|
| 1871–1879 | 147.0 | 108.9 | |
| 1880–1889 | 136.9 | 95.6 | |
| 1890–1899 | 137.7 | 99.6 | |
| 1900–1909 | 118.2 | 103.3 | |
| 1910–1919 | 92.0 | 117.7 | |
| 1920–1929 | 78.6 | 121.7 | |
| 1930–1939 | 90.4 | 115.8 | |
| 1940–1945 | 90.1 | 108.2 |
In the 1870s, 39.9% of everyone who died in New York was under two years old, and they died in the heat; by the 1940s, 7.6%. The inversion above is that group leaving the record — what is left is the year the adults always had.
CDC/NCHS weekly provisional deaths by jurisdiction of occurrence. United States, 2014-2019: 16,528,794 deaths.
Each dial is one year of American deaths from one cause, drawn as a clock. The thin circle is an ordinary week. Where the shape bulges outward, that cause kills more than usual; where it pulls in, less. A cause with no season is a circle.
Flu and pneumonia
110.9points
Chronic lung disease
41.4points
Cause unknown
33.7points
Blood infection
33.5points
Other lung disease
32.1points
Alzheimer's
29.8points
Diabetes
27.0points
Kidney disease
26.9points
Heart disease
24.9points
Stroke
19.3points
Cancer
6.8points
How to read a dial
The number is the gap between that cause's busiest week of the year and its quietest, as a share of an ordinary week. Flu and pneumonia runs 178.4 at its worst and 67.5 at its best. Cancer runs 104.3 and 97.5.
The spoke marks the busiest week. Every dial is drawn to the same scale.
How far the year moves each cause, in points
over 10030 to 10020 to 30under 20All eleven bulge in the same place. Whatever it is written on the certificate — the heart, the lungs, the kidneys, the brain — the deadliest weeks of the American year are the first two of January, and the safest are the last of August.
With one exception.
Each dot is one cause at its peak week, averaged over 2014–2019.
Ten of the eleven peak in week 1 or 2 — the first fortnight of January. Deaths with no cause written on the certificate peak in late December instead.
The odd one out is not a disease. It is the box on the certificate for a death nobody could assign a reason to. In the last week before Christmas those run 13.76 in every thousand deaths; in the second week of January, 10.94. In every one of the six years, late December is higher than early January.
Why is not in this record, and this page does not guess.
Take the holidays out of the calendar and what is left is a season. It has one peak, in the middle of September. Count back 266 days from that peak — the length of an average pregnancy — and you land on Christmas Day.
The two ends of that line have nothing to do with each other. September is full because of when people conceive. Christmas Day is empty for a different reason entirely — and the reason shows up the moment you ask what weekday it was.
Births on each weekday, averaged over 1994–2014. On the left, ordinary days. On the right, the 21 Christmas Days.
An ordinary day
Christmas Day
Both panels are drawn to the same scale. An ordinary Tuesday brings 12,708 babies and an ordinary Sunday 7,522 — a difference of 69%. Across the seven Christmas Days the difference is 7%.
However the calendar falls, Christmas Day brings about 6,601 babies — fewer than any Sunday of the year. That is why the hole looks so different from one year to the next: a Christmas that lands midweek runs 45.3% below an ordinary day, and one that lands at the weekend only 18.9%.
And the hole is getting deeper. Counting only the Christmases that fell on a weekday, the drop went from 41.9% in 1994–2003 to 49.0% in 2004–2014.
A June wedding is the oldest piece of calendar folklore in America. It was not always true, and it stopped being true. Before June it was October. By the end of the record it is September.
Every dot is one year, sitting on the month that had the most weddings a day that year.
In New York, June took the lead in the 1890s and then held it every single year from 1900 to 1937 — 38 years in a row with no exception at all. That is the strongest habit anywhere in this record.
It did not hold. In the last twelve years of the record — 1990 to 2001 — the busiest wedding month in Connecticut is September.
Both panels are drawn to one scale. 100 is an average day of the same year, so a month at 170 had seventy percent more weddings a day than usual.
New York City1866–1937
June takes it from October, and May empties out behind them.
Connecticut1900–2001
September climbs for eighty years and passes June at the end.
New York is monthly certificates and Connecticut is dated records, so the two are drawn separately and never joined into one line. Both are indexed within their own year, which takes out the growth of the city and the progress of the digitising.
The collapses are the quiet part. In New York, May fell from 109.9 to 78.7 while June was climbing — a month that used to be a wedding month and simply is not one any more.
That is the one wedding date anybody can guess, and the guess is right — but only if you were born after about 1960. Before the war, Valentine’s Day was an ordinary day in February.
1900–19194.1%
778 on the fourteenth
1920–19394.3%
801 on the fourteenth
1940–19597.9%
2,678 on the fourteenth
1960–19797.9%
2,717 on the fourteenth
1980–200122.0%
7,602 on the fourteenth
Every bar is that day’s share of all the marriages those Februaries held. All five Februaries are drawn to one scale, 0 to 24%.
a fair share of the month — 3.5%Against the other days of its own February — same year, same weekday — the fourteenth ran at 121.9% in 1900–1919. By 1980–2001 it ran at 839.1%. It is a tradition with a start date, which is a stranger thing than a tradition.
In those last years it is the single most-married date on the calendar: no day in June and no day in September comes near it.
Every dot is one date — not a 12 June, but the twelfth of June in one particular year. Dots that would land on each other stack up instead.
The February 14 1987 was a Saturday and the third-largest single wedding day in 102 years of the record. Every other day in the top twenty is in June or September.
In 1900 a Connecticut wedding could be on any day of the week, and most of them were not on a Saturday. Inside one lifetime, six of the seven days emptied out.
Every column is one year, divided into the seven days of the week. The whole column is that year’s weddings, so a band getting fatter means another one is getting thinner.
MondayTuesdayWednesdayThursdayFridaySaturdaySundayIn 1900–1939, 24% of Connecticut weddings were on a Saturday and 18.6% were on a Wednesday. By 1980–2001 it is 65.1% on a Saturday and 2.4% on a Wednesday.
The high-water mark is 1967, when 71.5% of every wedding in the state happened on the same day of the week. The only day that grew alongside it is Friday.
An almanac ends with a page correcting its own record. Two of the strongest patterns in this data are not facts about people. They are facts about how the writing-down was done — and one of them has something real hiding inside it.
Every December somebody writes that deaths spike at New Year. In a record of 3,115,245 deaths, more than a quarter of them fall on 1 January. Almost none of that is real.
Every record3,115,245 deaths, however the date was recorded
Only the records that carry a day2,242,443 deaths whose date includes a day of the month
Both strips are the same year and the same scale — an ordinary day is 6,229 deaths on the left and 6,124 on the right. The only thing the filter removes is dates that were never recorded.
Take those away and something is still there. Among the deaths that do carry a real day, 1 January runs at 192.5 — nearly twice an ordinary day. So do a handful of other dates: April 15 at 126.9, December 31 at 119.0, March 1 at 114.6 and February 1 at 114.0. Those are the days a person writes down when they are not sure.
Properly dated American records show no New Year needle at all — just a wide winter hump that peaks in the second week of January, which is the plate near the top of this page.
People are supposed to be more likely to die near their own birthday. Measured against what chance alone predicts, one day of it is a typing mistake and the rest of it is real.
Taken out
+73.2%
the birthday itself — more than ten times anything beside it, and a date typed twice looks exactly like this
Leave the birthday out and the week either side still runs +4.11% above chance. That is a real hump: it builds and decays over weeks, which a filing error cannot do.
And it is lopsided. The seven days after your birthday run +5.53%; the seven days before, +2.69%. Widen the window and the gap widens with it — at a month either side it is +2.18% against +0.47%. Past a month, both sides are flat at -0.5%, which is how you know the comparison is honest.
| Births | CDC/NCHS and SSA daily counts via the FiveThirtyEight archive. 1994–2014, 85,712,738 births. |
|---|---|
| Deaths, then | NYC Municipal Archives Index to Digitized Death Certificates. 1871–1945, 2,904,802 drawn of 3,139,673 usable. |
| Marriages | New York City Municipal Archives Index to Marriage License Indices, 1866–1937, 5,300,590 certificates, month only. Connecticut Vital Records Index of Marriages, 1900–2001, 2,540,428 marriages, exact dates. |
| Deaths, the errata | Wikidata humans with a recorded date of death, 1800–2020, via the QLever SPARQL endpoint. 3,115,245 deaths, each carrying the precision its date was recorded at, and 1,961,433 people with both dates to the day. |
| Deaths, now | CDC/NCHS weekly deaths by jurisdiction of occurrence. 2014-2019, 16,528,794 deaths. Complete years only; 2020 on is excluded so a pandemic is not compared with a pandemic-free baseline. |
All free and openly licensed. Extracts, licences and vintages: research/source.md.
Every number above is emitted by pipeline/15_site_payload.py and pipeline/16_calendar_payload.py, which assert each one against the project’s frozen claim record and fail the build on a drift.