Every Passport in the World Puts the Surname First

There is an appealing shortcut for the question "which order does this country write names in?" — look at the passport. It is issued by the state itself, it has captioned fields, and the state presumably knows how its own citizens are named.

We tried that shortcut across 64 locales. The passport answered surname first 64 times out of 64.

Not "mostly". All of them. The United States, Germany, France, Brazil, Iceland, Bulgaria — every one of them prints the family name above the given names on the data page, and every one writes the family name first in the machine-readable strip at the bottom. A test that returns the same value for every member of the set is not a test. It is a constant.

The honest answer, taken instead from how each country enumerates the parts of a name in its own language, is 15 of 64 — and one of those fifteen is a live argument inside the country concerned.

Three sources, three answers, and the one that looks most authoritative is the one carrying no information at all.

Where the constant comes from

The uniformity is not a coincidence and it is not cultural convergence. It is one document: ICAO Doc 9303, Machine Readable Travel Documents, maintained by the International Civil Aviation Organization and mandatory for any passport that is going to be read by a machine at a border.

Doc 9303 is careful not to say "surname". Its Part 3 instructs that the holder's full name

shall be divided where possible into two parts: the first representing that portion which the issuing State or organization identifies as the primary identifier (e.g. surname, maiden name plus married name, family name), and the second representing all remaining components which the issuing State or organization considers as collectively representing a secondary identifier (e.g. given names, initials).

Read that again, because it is the crux. ICAO does not legislate which component is the family name. It leaves the semantics entirely to the issuing state — a genuinely culture-neutral piece of drafting. What it does legislate is the order: primary identifier first, then two filler characters <<, then the secondary identifier. That part is not optional and not negotiable, because a border scanner has to know where to split the string.

So the machine-readable zone of a passport from anywhere on earth looks like this:

P<USASMITH<<JOHN<MICHAEL<<<<<<<<<<<<<<<<<<<<
P<HUNNAGY<<ISTVAN<<<<<<<<<<<<<<<<<<<<<<<<<<<
P<BGRGEORGIEV<<IVAN<PETROV<<<<<<<<<<<<<<<<<<

An American, a Hungarian and a Bulgarian. In life these three write their names in three different orders. In the MRZ they are identical in structure, and the family name leads in all three.

The visual zone has slightly more latitude — Doc 9303 permits either two separate captioned fields or a single combined field with a comma separating the identifiers — but the recommended layout places the primary identifier field above the secondary one, and in practice that is what states print. The captions are bilingual by rule: where the issuing state's official language is not English, French or Spanish, the printed caption must be followed by an oblique and the equivalent in one of those three. Hence Фамилия/Surname, Soyadı/Surname, 姓/Surname, نام خانوادگی/Surname.

The European Union bolted the same frame onto ID cards

Passports were only ever half the story, because inside the Schengen area most people cross borders on a national identity card. Regulation (EU) 2019/1157 closed that gap. Article 3(1) requires identity cards to be produced in ID-1 format with a machine-readable zone, "based on the specifications and minimum security standards set out in ICAO Document 9303". Article 3(2) goes further:

The data elements included on identity cards shall comply with the specifications set out in part 5 of ICAO document 9303.

Part 5 is the ICAO specification for credit-card-sized official travel documents. Its data element directory is what an EU identity card must now follow. The Regulation gives member states two derogations — the document number may sit in a different zone, and the gender field is optional — and neither of them touches the name block.

Old cards that fail the standard expire on a schedule: cards without a functional MRZ or below the Part 2 security minimum ceased to be valid by 3 August 2026, and everything else non-compliant dies by 3 August 2031. Which means that by 2031 the last European identity documents that could have printed a name in a national order will have been withdrawn.

Two standards bodies, one for aviation and one for the internal market, have between them guaranteed that no official identity document anywhere can tell you what order a name goes in. They did not do it carelessly — a fixed order is exactly what you want for machine reading. But it makes the document worthless as evidence about the culture.

The criterion that does carry information

If the form cannot tell you, ask the language. Every country that cares about name order has a phrase or an abbreviation for "the parts of a person's name", and that phrase lists them in the order the country actually thinks in.

Notice how much these phrases say that a form field cannot. Hungarian családi és utónév is a compound noun that would be ungrammatical in the other order. Chinese 姓名 and Japanese 氏名 are single words, not two fields — the order is welded into the vocabulary. Arabic الاسم الرباعي does not describe two components at all; it describes four, and its shape is a chain of ancestry, not a first/last pair.

And Slavic ПІБ / ФИО / ТАӘ are the sharpest illustration of the problem, because they are also surname-first — but for a completely different reason than a passport is. They are surname-first because that is how a Ukrainian, Russian or Kazakh office actually lists a person. The passport is surname-first because a scanner in Frankfurt needs a delimiter. The two facts look identical on paper and have nothing to do with each other.

Applying the language criterion across our 64 locales gives 15 surname-first: el_GR, en_UG, he_IL, hu_HU, hy_AM, ja_JP, kk_KZ, ko_KR, ro_MD, ru_RU, tr_TR, uk_UA, vi_VN, zh_CN, zh_TW. Forty-nine are given-name-first. The passport criterion would have flagged all 64 — more than four times the true figure.

Bulgaria: the card disagrees with the country

The most instructive case in the set is the one where the document and the language contradict each other outright, on the same person, in the same week.

Bulgarian law is explicit. Article 9 of the Закон за гражданската регистрация (Civil Registration Act), as amended in ДВ 96/2004, states:

Името на български гражданин, роден на територията на Република България, се състои от собствено, бащино и фамилно име. (The name of a Bulgarian citizen born on the territory of the Republic of Bulgaria consists of a given name, a father's name and a family name.)

Three parts, in that order, and the same article requires all three to be entered in the birth record. Article 13 then specifies how the middle part is built: the father's given name plus the suffix -ов or -ев, with the ending agreeing with the child's gender. Article 14 applies the identical rule one generation up to produce the family name — which is why in Bulgaria the surname and the patronymic are morphologically the same kind of word, and why the surname corpus doubles as a lookup table for patronymic formation.

So a Bulgarian named Ivan, whose father is Petar and whose family name is Georgiev, is:

Иван  Петров  Георгиев
 │      │        └── фамилно име   (family name, from the grandfather)
 │      └─────────── бащино име    (father's name, from the father)
 └────────────────── собствено име (given name)

Every school register, every utility bill, every voter roll, every time he signs anything, that is the string. The patronymic sits in the middle.

Now look at his identity card. The лична карта is an EU ID-1 card, so it follows Part 5 of Doc 9303, so the printed name block runs:

Exactly reversed, with the patronymic moved from the middle to the end. And the MRZ compresses it into GEORGIEV<<IVAN<PETROV — family name in the primary identifier, given name and patronymic packed together into the secondary identifier, because ICAO has two slots and Bulgaria has three components.

This is the single cleanest demonstration of the article's thesis. The document is not lying — it is filling in a standard form correctly. But the card contradicts the way a Bulgarian writes his own name everywhere except on the card, and anybody who used the card as ground truth would conclude that Bulgaria is a surname-first country. It is not. It is a given-name-first country whose middle component happens to be a patronymic.

The middle position is not a Bulgarian eccentricity either. In our data the patronymic sits between the given name and the family name in three locales — bg_BG, ar_JO and ar_SA. Saudi Arabia's الاسم الرباعي runs given name → father → grandfather → family name; Jordan's civil status law builds the same chain. In none of them does a First name / Last name pair have anywhere to put the middle element, and in none of them does the identity card print the components in the order the law states them.

There is one locale in our set that goes the other way and it is worth naming, because it is the exception that proves how narrow the ICAO rule's reach is. The Iranian کارت ملی هوشمند prints نام (given name) before نام خانوادگی (family name), and puts نام پدر (father's name) after the complete name rather than inside it — because in Iranian practice the father's name is a document attribute, like the date of birth, not a component of the name. Iran's national ID card is neither an ICAO travel document nor subject to EU 2019/1157. Its passport, of course, is ICAO-compliant, and its MRZ puts the surname first like everyone else's.

Uganda: when "as on the document" is itself the argument

Bulgaria is a clean contradiction. Uganda is a messier and more interesting one, because there the question of name order is unresolved, politically charged, and — according to at least one serious critic — an artefact of a database.

The surface facts point one way. Registration forms and verification services from Uganda's National Identification and Registration Authority ask for surname first, then given name. School certificates print surname first. People introduce themselves surname first: "Nakalinzi Maureen", "Sejjaaka Samuel". If your criterion is "how does the paperwork list it", Uganda is unambiguously a surname-first country.

Prof. Samuel Sejjaaka, a Ugandan academic, argued in a two-part Daily Monitor commentary that this is not tradition but an accident. His account: when national examinations were computerised, candidates were listed alphabetically by surname then first name — a sort order, not a naming convention. When the examinations board came to print transcripts and certificates, it took the sort order for the name and never let go. The result is that Ugandans who left school as recently as 2000 write themselves as "Karuhanga James Mugisha". Sejjaaka's own recollection is that introducing yourself surname-first "was not a done thing" a generation earlier.

The follow-up column recorded the obvious rebuttal — that if Ugandans are comfortable with surname-first, no Heathrow immigration officer and no search engine gets a vote — and Sejjaaka conceded there was more to it than he had assumed. The debate is not settled and we are not settling it here.

What makes Uganda worth a section is what happened in 2020. Ahead of the 2021 elections, Uganda's Electoral Commission announced that candidates whose names differed from those on their academic documents would not be nominated. Because the school certificates carried one order and everyday life carried another, more than 160 politicians swore deed polls to rearrange their own names, including Robert Kyagulanyi (Bobi Wine), Henry Tumukunde and Mugisha Muntu. On 6 October 2020 the president did the same: the deed poll declares that he had been born Yoweri Tibuhaburwa Kaguta, that his certificates read Yoweri Tibuhaburwa Museveni, and that the proper legal order of his name is henceforth Yoweri Tibuhaburwa Kaguta Museveni — given name first. He formally renounced the three variants he had been using, one of which was the name the world knew him by.

Set that next to the passport question. In Uganda, "which order is correct" was decided by an examinations database in the 1990s, contested by academics in the 2010s, and then had to be settled one person at a time by sworn instrument in 2020 because an election commission required documents to agree with each other. The document did not record the convention. The document created the convention, and then people had to go to a lawyer to escape it.

Our own data marks Uganda as surname-first while flagging the row as an estimate rather than a fact — which is the only defensible position given that the country has not agreed with itself.

CLDR will not rescue you either

At this point the instinct of any developer is to look for a machine-readable list. There is one, and it is maintained by Unicode: the CLDR person-name data introduced in CLDR 42 contains exactly the element you want:

<nameOrderLocales order="givenFirst">und en</nameOrderLocales>
<nameOrderLocales order="surnameFirst">ja ko vi yue zh</nameOrderLocales>

This is a trap, and it is a subtle one, because the element is doing something genuinely useful — just not the thing its name suggests.

nameOrderLocales is not a list of countries that write the surname first. It is a list attached to a display locale, telling a formatter: "when you are rendering text in me, and the name you were handed belongs to one of these locales, put the surname first." The spec says so directly — the element "contains information about selecting patterns based on the locale of a passed in PersonName object". The und entry "matches all base locales not explicitly listed", and it may appear in exactly one of the two lists, never both.

The consequence is that the list is different in every file, because it encodes an editorial convention about how this language's readers expect foreign names to appear. Here is the actual data, read out of the CLDR 47 release:

Read the German row against the English row. German CLDR data says a Japanese name should be rendered given-name-first; English CLDR data says surname-first. Same person, same name, two Unicode-blessed answers, differing on the language of the surrounding paragraph. Neither is wrong, because the question CLDR is answering is "how do my readers expect to see this name?" — and German and English readers genuinely differ.

The English list gained ja for a documented reason. In May 2019 Japan's foreign and education ministers proposed that Japanese names be rendered surname-first in the Roman alphabet; the government adopted it in September and began applying it to official documents on 1 January 2020, with the family name optionally capitalised to mark it — ABE Shinzo, KONO Taro. CLDR has a matching option, surnameFirstAllCaps, for precisely this. So the English list encodes a 2019 diplomatic request, and the German list has not yet acted on it. That is a fine thing for a rendering library to encode. It is not a fact about Japan.

Now check the list against reality. Applied to our 64 locales, the English surnameFirst list matches five: ja_JP, ko_KR, vi_VN, zh_CN, zh_TW. The language criterion says fifteen. The ten it misses are:

el_GR, en_UG, he_IL, hu_HU, hy_AM, kk_KZ, ro_MD, ru_RU, tr_TR, uk_UA

Hungary is in that list. Hungarian is routinely described as the only European language whose native order is family name first, its own statute calls the field családi és utónév, and the English CLDR list does not contain hu — while the Hungarian, Japanese and Korean lists all do. Ukraine, Russia and Kazakhstan are in that list, and their entire administrative vocabulary is built on surname-first abbreviations.

And CLDR is right about Bulgaria in a way the ID card is not: bg appears in the givenFirst list, which matches Article 9 exactly. So on the one locale where the passport and the law contradict each other, CLDR sides with the law — and is still useless as a country list, because it only has 4 to 11 language subtags in it depending on which file you opened.

There is one more limit worth knowing. CLDR's PersonName object has fields title, given, given2, surname, surname2, generation, credentials. A Bulgarian patronymic has to go in given2 — the field documented as "additional given name or names or middle name" and explicitly noted as covering "patronymic names in Slavic regions". That works structurally, but it means the model records a Bulgarian бащино име as a kind of middle name, and a Spanish second surname as surname2, and offers no way to say "this component is derived from the father's given name and its suffix agrees with the child's gender". CLDR is a formatting library. It was never trying to be a registry of naming law.

Where the question has no answer at all

Two locales in our set are worth mentioning because they break the premise rather than the criterion.

Iceland. Icelandic law — Lög um mannanöfn nr. 45/1996 — does not have a surname field in the sense the rest of this article assumes. The register (Þjóðskrá Íslands) keeps eiginnafn, millinafn and kenninafn, and the kenninafn is normally a patronymic: the father's or mother's name in the genitive plus -son or -dóttir (with the gender-neutral -bur added by the self-determination act nr. 80/2019). Our Icelandic corpus reflects this directly — 284 patronymics in each of the male and female surname files, mirrored pairs like Jónsson/Jónsdóttir and Sigurðsson/Sigurðardóttir carrying identical weights, plus 59 genuinely inherited family names (Blöndal, Briem, Thoroddsen, Nordal) that are the same for both sexes. Asking whether Iceland puts the surname first is asking about a component that mostly is not a surname.

Indonesia. The KTP-el identity card has a single name field, Nama Lengkap — full name, one string. Indonesian law (UU 23/2006 as amended by UU 24/2013, with the recording requirements in Permendagri 73/2022) does not create a family name at all. Mononyms were normal at the highest level — Sukarno, Suharto — and although newborns have been required to have at least two words since 2022, every adult mononym stays valid for life. Patronymic markers exist (bin, binti, putra, putri) but they live inside that one string and have no field of their own. There is no order to get right, because there is only one box.

Indonesia also collides with ICAO in a documented way. Doc 9303 tells issuing states that a holder with a single name should have it written into the primary identifier, leaving the secondary identifier empty. ICAO's own Technical Advisory Group has recorded that this "has not been uniformly implemented" because some issuing authorities are unfamiliar with mononymous names, and that the resulting non-compliance causes travellers unnecessary delays — and tasked its working group with reviewing the primary/secondary identifier conventions for exactly that reason. The standard that made every passport look the same still cannot cope with a person who has one name.

What to do instead

1. Never derive name order from a document layout. ICAO Doc 9303 and Regulation (EU) 2019/1157 have made every machine-readable identity document in the world agree, so agreement tells you nothing. If your source of truth is a passport specimen, your answer is "surname first" for every country on earth and you will never notice the error.
2. Derive it from how the country enumerates the parts in its own language. ПІБ, ФИО, ТАӘ, 氏名, 姓名, 성명, családi és utónév, Họ và tên, собствено, бащино и фамилно име. These phrases are grammar, not layout, and grammar does not get reordered to suit a scanner.
3. Do not treat CLDR nameOrderLocales as a country list. It is a per-display-locale rendering convention with between four and eleven language subtags in it. It is correct for what it does and it will silently mis-answer the question you are asking.
4. Allow the patronymic to be in the middle. Three of our locales put it there, and both the ICAO two-slot model and the First name / Last name pair force it somewhere else.
5. Check whether the country has the component at all. Iceland's "surname" is a patronymic. Indonesia has one undivided name field. Uganda's surname is a clan name deliberately chosen so that it differs from the father's.

And the practical consequence, which is the reason any of this matters outside a data set:

A form with First name and Last name has already made the decision for the user. It has fixed the count at two, fixed the order at given-then-family, and assumed that both concepts exist. For roughly three quarters of the locales we ship that is a reasonable default. For Bulgaria it drops the middle component or misfiles it. For Saudi Arabia it collapses a four-link chain into two boxes. For Indonesia it splits a string the law does not permit to be split. For Iceland it labels a patronymic as a family name. And for Uganda it picks a side in an argument the country is still having.

The passport will not warn you about any of this, because the passport was designed by people whose only requirement was that a machine in another country could find the delimiter.

Data and sources

Data as of 2026-07-22. Counts were computed over our 64-locale set. Surname-first by the language criterion: 15 (el_GR, en_UG, he_IL, hu_HU, hy_AM, ja_JP, kk_KZ, ko_KR, ro_MD, ru_RU, tr_TR, uk_UA, vi_VN, zh_CN, zh_TW); 14 if the contested Ugandan row is excluded. Given-name-first: 49. Patronymic in the middle position: bg_BG, ar_JO, ar_SA. Match against the CLDR 47 English surnameFirst list: 5 locales, missing 10.

Standards

• ICAO Doc 9303, Machine Readable Travel Documents, Part 3 — Specifications Common to all MRTDs. Definition of primary and secondary identifier; the rule that the primary identifier is written first in the MRZ followed by two < fillers; transliteration and truncation rules; caption language requirement. Part 4 covers passports (39 name positions in the TD3 MRZ), Part 5 the ID-1 card (30 positions).
• ICAO TAG/TRIP working paper on naming conventions — records that Part 4 §4.1.1 requires a mononym to be placed in the primary identifier field with the secondary left blank, that this "has not been uniformly implemented", and that the NTWG was tasked with reviewing primary/secondary identifier representation in the VIZ and MRZ.
• Regulation (EU) 2019/1157 on strengthening the security of identity cards. Article 3(1): ID-1 format, MRZ, ICAO Doc 9303 minimum security standards. Article 3(2): data elements per Part 5 of Doc 9303, with derogations only for document number placement and optional gender. Article 5: validity cut-offs of 3 August 2026 and 3 August 2031. The Regulation's fixed reference is to the seventh (2015) edition of Doc 9303.
• Unicode LDML Part 8: Person Names (UTS #35). nameOrderLocales semantics, the und fallback rule, the field inventory (title, given, given2, surname, surname2, generation, credentials), and surnameFirstAllCaps. Introduced in CLDR 42.
• CLDR data quoted from the cldr-json distribution, release 47.0.0, files cldr-person-names-full/main/<locale>/personNames.json, read 2026-07-22.

Bulgaria

• Закон за гражданската регистрация, чл. 9 (изм. ДВ бр. 96/2004) — three-part name; чл. 13 — formation of the бащино име from the father's given name with the -ов/-ев suffix agreeing with the child's gender; чл. 14 — the same rule one generation up for the family name.
• Card field order (Фамилия/Surname → Име/Name → Презиме/Father's name) and MRZ encoding GEORGIEV<<IVAN<PETROV follow from Part 5 of Doc 9303 as applied by Regulation 2019/1157.

Uganda

• Samuel Sejjaaka, "What order of your name(s) says about state of our education system", Daily Monitor, Parts I and II. The computerised-examinations argument and the rebuttal.
• Museveni renames self Tibuhaburwa, Daily Monitor, October 2020. Deed poll sworn 6 October 2020; the declared order Yoweri Tibuhaburwa Kaguta Museveni; the renunciation of three earlier variants; the Electoral Commission requirement that names match academic documents; more than 160 politicians swearing deed polls to reorder their names.
• Ugandan surnames are clan names drawn from a clan's own name stock rather than from the father — M. B. Nsimbi, Baganda Traditional Personal Names (Uganda Journal, 1950); J. Roscoe, The Baganda (1911). This is the opposite of a patronymic system and is why the Ugandan birth register has name(s) and surname fields but no patronymic field.

Japan

• The May 2019 proposal by foreign minister Kōno Tarō and education minister Shibayama Masahiko; adoption in September 2019; application to official Roman-alphabet documents from 1 January 2020, with capitalisation of the family name. A December 2000 recommendation by the Japanese Language Council to the same effect had not taken hold.

Other locales

• Iceland — Lög um mannanöfn nr. 45/1996, arts. 8–9 (kenninafn = parent's name in the genitive + -son/-dóttir); lög um kynrænt sjálfræði nr. 80/2019 (-bur); Þjóðskrá Íslands register fields eiginnafn / millinafn / kenninafn. Corpus figures: 284 patronymics per gender file plus 59 inherited family names shared by both.
• Indonesia — UU No. 23/2006 on population administration as amended by UU No. 24/2013; Permendagri No. 73/2022 on name recording (minimum two words for newborns); the KTP-el single Nama Lengkap field.
• Hungary — 414/2015. (XII. 23.) Korm. rendelet on the issue of the identity card, which uses the statutory phrase családi és utónév; the card's data content is fixed by 1992. évi LXVI. törvény.
• Iran — سازمان ثبت احوال کشور; the smart national card records نام, نام خانوادگی, نام پدر in that order, with the father's name as a document attribute rather than a component of the name.
• Kazakhstan — ТАӘ = Тегі, Аты, Әкесінің аты, the Kazakh equivalent of ФИО, used throughout official forms and payment details.

A note on what we could not verify. ICAO's visual-inspection-zone layout is a recommendation with a permitted single-field alternative, so "the surname is printed above the given names" is an observation about how states have implemented it, not a quotation from the standard. The Ugandan case is explicitly an estimate in our data rather than a settled fact, and we have marked it as such rather than let a disputed row look like a documented one.