Why We Have Leap Years (and Why 2100 Won't Be One)
Print the same 365 boxes every year and the seasons will not stay inside them.
That is why we have leap years: Earth's cycle of seasons lasts about 365.2422 days, so a 365-day calendar finishes almost six hours too soon. The spare fraction is saved and periodically returned as a whole day. The Gregorian calendar usually does that every fourth year, skips century years, then restores centuries divisible by 400. That last turn is why 2000 had February 29 while 1900 did not and 2100 will not.
It looks like an untidy exception. It is actually a carefully reduced error.
Why do we have leap years? Run the calendar without them
Start with an ordinary 365-day year and leave the fraction unpaid:
365.2422 − 365 = 0.2422 day per year.
After four years, the calendar is about 0.9688 day ahead of the seasonal cycle. After one hundred years, it is about 24.22 days ahead. A date called March 20 would arrive while the seasonal position once associated with that date was still more than three weeks away. The U.S. Naval Observatory gives the northern spring example plainly: without correction, a March 20 beginning of spring would move to about April 13 after a century.
The Sun has not slipped. Your labels have.
Adding one day every four years nearly clears the account: four lots of 0.2422 make just under one day. That was the working idea of the Julian calendar. Its average year was 365.25 days, however, about eleven minutes longer than the tropical year. Small on a clock, large across centuries. By the sixteenth century the March equinox was falling around March 11 on the Julian calendar, about ten days before the March 21 reference restored in the Gregorian reform. The story of the 1582 reform belongs in its own article; the arithmetic that forced it is already visible here.
The leap year rules explained in three lines
The Gregorian test works from broad rule to narrow exception:
- If the year is not divisible by 4, it is a common year.
- If it is divisible by 4, it is a leap year — unless it is also divisible by 100.
- A century year becomes a leap year again when it is divisible by 400.
Try the years people trip over. The year 2028 divides by 4 and is not a century, so it gets February 29. The year 2100 divides by both 4 and 100 but not 400, so it does not. The year 2000 passes all three tests and keeps the extra day. So will 2400.
Across 400 years, this makes 97 leap years rather than the Julian calendar's 100. The average is therefore:
365 + 97 ÷ 400 = 365.2425 days.
That is only about 0.0003 day longer than the approximately 365.2422-day tropical year. The Naval Observatory estimates that the difference takes about 3,300 years to build to one day. You may encounter a more exact-sounding figure such as 3,030 years, but it depends on which value of the changing tropical year is used. “About 3,300” is the official estimate, and pretending the sky supplies a permanent decimal would be false precision.
Could the rule be refined? Yes. Calendar reform proposals have suggested omitting another leap day on a much longer cycle. No such repair is urgent, and future keepers would also need to account for the fact that Earth's rotation and the measured seasonal year do not remain perfectly constant. The present rule is not eternal mathematics. It is excellent civil engineering.
Why February 29 is the leap day
February did not win a contest to become the odd month. It inherited the place where Rome already made its calendar adjustment.
Romans identified dates by counting backward from fixed points in a month. In the Julian system, the leap adjustment duplicated the “sixth day before the Kalends of March,” a date near what we call February 24. That repeated sixth day gave the Latin description bissextile — twice-sixth — still used for leap years in several languages. The Royal Observatory of Belgium notes that, under that older way of counting, February 24 or 25 could be treated as the inserted day.
Modern calendars instead display one clean sequence, February 1 through 29, so the visible extra date sits at the end. February 29 is therefore the modern label for an adjustment whose address is Roman. It is not there because Augustus stole a day for August, a popular story that does not fit the surviving calendar history.
The extra day also explains why dates after February can change weekday by two steps between one year and the next. A normal 365-day year is 52 weeks plus one day. A leap year is 52 weeks plus two.
A February 29 birthday meets the law, not astronomy
A person born on February 29 has one birthday every year in the ordinary sense: their age advances continuously, not in four-year jumps. The awkward question is which civil date counts as the anniversary when a statute, contract, licence or age threshold needs one.
There is no worldwide answer. Local law supplies it.
For one concrete example, section 6 of Scotland's Age of Legal Capacity Act 1991 treats March 1 as the relevant anniversary in a non-leap year. Another jurisdiction, statute or contract may frame a deadline differently. That is why a sweeping claim that every leap-day baby “legally ages on February 28” or “legally ages on March 1” is unsafe. For an actual deadline, read the governing rule rather than an astronomy explainer.
Celebration is easier. Some choose February 28, some March 1, and some take both.
A different place to put the remainder
Every calendar has to disclose where its fractions go, even if the disclosure is buried in the rules. The Gregorian calendar inserts a counted date inside February. Day 4's 364-day reckoning uses a different answer: it holds the remainder after counted day 364 in one or two uncounted days called the Waiting Space, then begins again from the March equinox supplied by its canonical year table.
That means no counted year-day shifts to another weekday. The year is always 52 whole weeks. It also means those waiting days do not belong to the sacred week at all — a cost and a design choice, not a hidden Gregorian leap day under another name.
Day 4 shows that sacred count beside the everyday calendar. It does not replace civil leap-year rules, and you should still expect February 29 on your phone in 2028 and no such date in 2100. The comparison is useful because it exposes the choice: insert the remainder inside the year, or hold it at the boundary.
Common questions
Why do we add a day every four years?
Because the seasonal year is about 365.2422 days, almost a quarter-day longer than a 365-day calendar. Four of those fractions total about 0.9688 day, so adding one whole day every four years keeps dates near the same seasons. “Every four years” is the first rule, not the whole rule, because the approximation adds slightly too much.
Why was 2000 a leap year but 1900 and 2100 are not?
All three are divisible by 4 and 100, which would normally make a century year common. The final Gregorian exception restores years divisible by 400. The year 2000 divides evenly by 400, so it was a leap year. Neither 1900 nor 2100 does, so both are common years. The same rule makes 2400 a leap year.
What would happen to the seasons if we stopped?
The calendar dates would move away from the seasonal cycle by about 0.2422 day each year, or roughly 24 days per century. Spring would not physically arrive later; the date printed beside it would. Over centuries, familiar seasonal months would trade places until northern spring dates eventually appeared in what the calendar called winter.
How accurate is the rule, and will it ever need fixing again?
The Gregorian average is 365.2425 days, close to the roughly 365.2422-day tropical year. The remaining difference accumulates to about one day in 3,300 years by the U.S. Naval Observatory's estimate. A future adjustment is mathematically possible, but no presently adopted global rule schedules it, and the slowly changing astronomical quantities make an exact forever-fix unrealistic.
Sources:
- U.S. Naval Observatory — Leap Years
- Explanatory Supplement to the Astronomical Almanac, chapter 15
- Royal Observatory of Belgium — The Roman and Julian Calendars
- Age of Legal Capacity (Scotland) Act 1991, section 6
The useful question is not whether a calendar contains an exception. It is whether you can see what the exception pays for. When 2100 arrives without February 29, look past the empty square: three omitted days in every four centuries are what keep March in spring.
