Methodology

Holiday sites disagree with each other constantly, and almost never because one of them has better information. They disagree because they count differently and do not say so. This page says so.

What counts as a national public holiday

A day off that applies across the whole country, granted by national law or by a national announcement with legal effect. That is the whole test, and it is applied the same way to every country.

What is therefore excluded, deliberately:

A handful of countries do not have one national calendar at all. Their hubs say that above the table instead of presenting a partial list as a complete one.

How a moved holiday is counted

Once, on the day it is observed. When a holiday falls on a weekend and the day off is granted on a nearby weekday, that is one holiday with a shifted day off, not two holidays.

This is the single largest source of disagreement between holiday sites, and the source data makes it easy to get wrong: the dataset this site builds on emits the weekend date and its replacement as two separate entries. Adding them up inflates every total. GOV.UK publishes 27 and 28 December 2027, not 25 to 28. The US Office of Personnel Management publishes eleven federal holidays, not fifteen. Counting once is what makes the totals here match the official ones, and it is why they are sometimes lower than elsewhere.

In the tables both dates are shown: the day off, with the calendar date it replaces beside it. In the JSON they are separate fields.

Which source wins

In this order, highest first:

  1. The official gazette or announcement for that specific year.A ministry publishing next year's dates settles them, and nothing computed overrides it.
  2. The statute. Where the calendar is set in standing law, the law is the source and no annual announcement is needed.
  3. The open source dataset. date-holidays, which supplies the rule for every holiday not covered above.
  4. Computation from the rule in force. Used only when nothing above settles the year, and labelled as such on the page.

35 corrections currently sit between levels two and three, across 15 countries. Every one names its instrument and the date it was checked, in the corrections log.

How dates are computed, and why nothing is typed in

No date on this site is entered by hand except a gazetted one. Each holiday carries a rule (a fixed date, a named weekday of a month, an offset from Easter, a Hijri month, a lunisolar date, an equinox), and every year is computed from that rule. Corrections carry a rule in the same grammar and go through the same engine. This matters because a hand-typed calendar is right the year it is written and quietly wrong afterwards.

It also means some dates genuinely cannot be known in advance. A Hijri date depends on a moon sighting and can move a day in either direction. Where the dataset has no date at all for a future year, the holiday is listed with the date marked as not yet announced, rather than dropped so the total quietly falls.

Gazetted, statutory, computed

Every country and year carries exactly one status:

Every status sentence on the site is generated from that value. There are no hand-written status claims left in the templates, because a hand-written one stops being true and nobody notices.

Which years are published, and which are not

2025, 2026 and 2027: the current year, one back and one forward. The window rolls forward automatically, and a scheduled rebuild keeps it moving without anyone remembering to deploy.

Two years ahead is deliberately not published. No authority anywhere fixes a calendar two full years out. China publishes each November for the following year, Spain re-fixes by BOE resolution each autumn, Japan's equinox dates come from the national observatory a year ahead, Argentina's bridge days come from an annual decree, and every Hijri date waits on a sighting. A second forward year would not be a calendar, it would be a projection presented as one, and the underlying dataset silently drops lunar holidays once its stored dates run out. Competitors publish it. That is not a reason to.

A year that has passed keeps its page. People look back at last year's calendar for real reasons, and deleting it would break every link to it. Those pages carry a note that the year is over, links to the current and next year, and figures that only a completed year can have.

Reporting an error

Send it through the contact form. A report gets checked against the official source rather than against another holiday site. If the date here is wrong it is corrected and the correction is published in the log with its source, so you can see what changed and on what authority. If the error is in the upstream dataset rather than in our layer, it is reported there too, because fixing it once at the source fixes it for everyone downstream.

If the report turns out to be a definition difference rather than an error, you get told that, with the statute. That is not a brush-off: it is the same answer we would want.

Who maintains this

HolidayGrid is built and maintained by FusionStudios, a Dutch digital agency. There is no editorial board and no claim of one. What there is instead is a published source for every departure from the underlying data, a date on every check, and a route to tell us we are wrong.

What this site does not do

Always confirm official dates with a local authority before booking travel or committing to anything contractual. That applies to every holiday calendar on the internet, including this one, and the 2027 pages in particular.