74 festivals filed · 74 with full entries · 1 with a sourced attendance figure
Country pass 07 · 74 entries, Q-IN-001 to Q-IN-074
Every country so far has contributed one thing. France schedules; Italy inherits; Spain is governed by Easter; Switzerland made dates conditional on the weather; the Emirates made them conditional on the sight of the moon; Japan produced one festival with two correct dates. India does something none of them did: it runs six calendars at the same time, and two festivals in the same week can be set by two different ones.
Holi is the Hindu lunisolar calendar. Pongal is the Tamil solar calendar. Thrissur Pooram is an asterism in a Malayalam solar month. Poila Boishakh is the Bengali year. Eid is the Hijri. Losar is the Tibetan. The Goa carnival is the Gregorian computus, left behind by the Portuguese. And the Kumbh Mela is set by the position of Jupiter, which is a ninth thing again.
So every record in this file names the calendar that governs it. That field did not exist before this pass and no other country has needed it.
The clearest demonstration is the second week of April, when five new years fall within a few days of each other on four different calendars: Baisakhi in Punjab, Puthandu in Tamil Nadu, Vishu in Kerala, Poila Boishakh in Bengal, Bohag Bihu in Assam. None of them is the Indian new year. Neither is Gudi Padwa, which the Deccan keeps a fortnight earlier on a fifth reckoning.
This atlas computes dates. It does Gregorian and Orthodox Easter, weekday rules, the tabular Hijri, seasonal windows, cycles, and — since Japan — the sexagenary day cycle, which lets it tell you that Tori no Ichi falls three times in November 2027.
It cannot compute the Hindu lunisolar calendar, and it does not try.
Twenty-eight of these seventy-four entries are governed by that calendar and not one of them gets a computed day. Fixing a date in it requires the true longitudes of the sun and the moon, an intercalary month inserted on an irregular rule, a choice between the purnimanta and amanta reckonings that differ by a fortnight, and a tithi that can begin in the middle of one day and end in the middle of the next. An approximate ephemeris does not give an approximate festival; it gives a plausible wrong date, which is worse here than no date at all.
Twenty-one of them now carry actual dates for 2026–2030 anyway — read off published tables, not computed here, and marked as read rather than calculated. Ten of those dates have two published values a day apart and both are recorded; twenty-five rest on a single source and say so. The reckoning most Indian readers actually use, drikpanchang, was refused at our network gateway, so the disagreements stand unsettled.
The 2027 disagreements are worth one more sentence because they are not random. Holi, Holika Dahan and Mahavir Jayanti are each exactly one day later in the panchang source than in the others, all within Phalguna and Chaitra. A whole cluster off by one is the signature of a tithi ending very close to sunrise, with two reckonings applying different rules to it — not of anyone making a mistake.
The rest carry a month window, the rule in words, and no computed day. Fifty-six of the seventy-four get no exact date from this engine at all, once the seasonal windows and the cycles are counted in. A reader planning Diwali in 2029 is told that it falls on the new moon of Kartika, in October or November, and is sent to an almanac. That is the honest output and it is what the file contains.
The Hindu lunisolar year is kept in step with the sun by inserting an extra month — the adhik maas — every two or three years. In this five-year window it happens twice: on 17 May 2026 and on 16 March 2029, and in no other year.
Everything after an insertion runs late. 2026 therefore carries the latest Diwali of the five years, on 8 November, and its Rath Yatra, Janmashtami, Ganesh Chaturthi and Navratri are all pushed back with it. 2029 does something stranger: the year splits in half. Holi on 1 March falls before the intercalary month and is early; everything after mid-March is late. A reader who learns the shape of one Indian year and assumes the next will resemble it is wrong in a way that has a mechanism.
By contrast 2028 has no insertion at all, and it shows: Dussehra falls in September that year, on the 27th, and Diwali on 17 October.
Navratri means nine nights, and the published dates say it runs nine civil days in 2026 and 2027 and eight in 2028, 2029 and 2030.
The cause is a kshaya tithi — a lunar day that begins and ends between two sunrises, and so never owns a calendar date at all. Two of the nine merge.
This entry previously carried a duration of nine days. It no longer carries a duration at all, because the only honest answer is the dates themselves. Any register that stores this festival as start plus nine produces a wrong end date in three of these five years, and ours did.
Two kinds of Indian date are computed. The Hijri entries — the two Eids and Muharram — come out of the tabular calendar, marked as tabular because the observed day is settled by sighting the moon and can move by one. And the solar dates are exact: Makar Sankranti, Baisakhi, Vishu, Puthandu sit where the sun puts them.
That second group deserves a note of its own. Makar Sankranti is the only date in this atlas whose drift is measured in centuries. The sidereal reckoning that fixes it moves against the Gregorian year by about a day every hundred years. It fell in late December in the fifth century; it is on 14 January now and moving to the 15th.
Eid al-Fitr, Eid al-Adha and Muharram drift about eleven days earlier each Gregorian year and pass through all twelve months in roughly thirty-three. The cards for them say DRIFTS rather than naming a month, as the Emirates entries do.
They are in this file for a reason worth stating plainly. India has one of the largest Muslim populations of any country, and a festival atlas that treats an Indian page as a Hindu calendar with a few additions has misdescribed the country. The Hijri is not an exception here. It is one of the six.
The Kumbh Mela at Prayagraj in 2025 was announced at more than 660 million.
We do not record it, and the reasons are the whole argument of this atlas.
It is a count of bathing events, not of people: a pilgrim who bathed on four of the principal days is in it four times. It was assembled from manual counts, train arrivals, satellite imagery and camera footage across forty-five days. No method was published. In the sources we reached, no issuing body is named. The Economist observed that the figure implies half of India attended.
The death toll from the crush of 29 January is disputed in the same way — about 30 officially, at least 79 by a count from hospital and police records, with opposition parties saying 1,500 were unaccounted for a week later. And the claim that 2025 was a once-in-144-years alignment is contested; we do not assert it.
The entry says all of that and prints no attendance figure. A number this large with no basis is not information about a festival. It is information about who wanted it published.
India answers the passport question differently from every country before it, and the answer is not a missing dataset.
In 2024 there were 9,951,722 foreign tourist arrivals and 2,948.19 million domestic tourist visits — a ratio of about 296 to 1. A foreign-origin layer for an Indian festival would describe roughly a third of one per cent of the movement.
Those two figures were read in the Ministry's own India Tourism Data Compendium 2025 on 11 November 2026, and reading it changed them. This section stood on 9.24 million arrivals for 2023, which is what the Press Information Bureau circulated in December 2024; the compendium prints 95,20,928 for that same year. We are not able to say which step of the Ministry's process moved the number, and the honest response is not to pick: the section now carries the year the Ministry currently reports and the day we read it. The same document is not internally consistent either — its text gives 95,20,928 arrivals for 2023 and its table 2.1.5 gives 95,20,924.
There is also a structural gap of the familiar kind: foreign arrivals are published monthly and by port of entry, domestic visits by state, and nobody publishes the intersection. A visitor landing in Delhi and travelling to Varanasi is counted at Delhi. That is the third country in a row — after Dubai and Spain — where two useful series exist and their intersection does not.
But the ratio matters more than the gap, and one festival proves it at the gate.
The Hornbill Festival in Nagaland publishes the only gate-level origin breakdown we have found for any Indian festival, and it is exact: 205,968 visitors across ten days of the 25th edition, of whom 147,224 were from Nagaland itself, 56,217 from the rest of India, and 2,527 were foreign. The daily counts are published and sum to the total precisely, so these are visits rather than unique people — and the shares are 71.5%, 27.3% and 1.2%.
That is the most internationally marketed festival in north-east India. One visitor in eighty is foreign.
Several of the largest things in this file cannot be attended anywhere.
Raksha Bandhan is a thread tied on a wrist. Diwali is lamps in households. Neither has a gate, an organiser or a crowd to count, and neither will ever carry an attendance figure. What they have instead is a measurable effect on the railways: both are among the heaviest travel periods of the Indian year, because the rite requires people to be in the same room and the families are scattered. Chhath empties Delhi and Mumbai of its eastern diaspora for a fortnight.
A reference work organised around venues handles this badly, and the honest record is the one that says the festival is everywhere, that it is domestic, and that its visible consequence is a transport problem.
Three entries are not annual, and two of them are the largest gatherings on earth. The Kumbh rotates between four river sites on a cycle set by Jupiter, roughly twelve years at each; the Ardh Kumbh runs on the six-year half-cycle at two of the four and not at the other two; and the Mahamastakabhisheka at Shravanabelagola anoints its monolith once every twelve years — last in 2018, next in 2030, and in the eleven years between there is a statue on a hill and no festival.
That is now five of seven countries in which a cycle had been filed as an annual event somewhere. It remains the most common factual error this atlas finds.
Seventy-four entries, Q-IN-001 to Q-IN-074. Twenty-six religious, seventeen seasonal, fourteen cultural, seven unique, four music, three carnival, two arts, one food. By calendar: twenty-eight Hindu lunisolar, eighteen Gregorian, eight Tamil solar, six Malayalam, four Tibetan, three Nanakshahi, three Hijri, two set by Jupiter, and two Bengali.
One attendance figure, one origin block, one refusal and two figures recorded as unusable with the numbers stated and the reasons given. That is a thin attendance layer by the standards of the European passes, and it is not a shortfall of effort: India's festivals are counted, where they are counted at all, by police estimate and administrative claim, and the two largest published numbers in the country turn out to be a count of dips and an economy.
Country-level — belongs to no festival on this page
This total is the two published series ADDED, and the addition is ours. The Ministry publishes them separately and never sums them, because a domestic VISIT and a foreign ARRIVAL are different units: one person on one trip produces several of the first and exactly one of the second. The total exists only as the base the two shares are computed against. Its last five digits are an artefact of the addition and carry no precision — domestic visits are published to 0.01 million, arrivals to the person, so this sum is meaningful to about four digits.
This cannot be attached to any festival. India publishes foreign arrivals monthly and by port of entry, and domestic visits by state. It does not publish their intersection, so foreign visitors by state by month — the series that would let a festival be described — does not exist. A tourist landing at Delhi and travelling to Varanasi is counted at Delhi. But the deeper point is that the passport question is close to the wrong question here: there were 2,948.19 million domestic tourist visits in 2024 against 9,951,722 foreign arrivals, a ratio of about 296 to 1. A foreign-origin layer for an Indian festival would describe roughly a third of one per cent of the movement. The one gate-level breakdown we found says the same thing from the other end: at the Hornbill Festival, the most internationally marketed event in the north-east, foreign visitors were 1.2% of attendance and 71% came from within Nagaland.
Published nearby and not visitor origin:
These recur on a lunar or other non-Gregorian calendar and pass through every month of the year. No month heading can hold them.