Male in one country, female in another: 47 names from birth records

47 names are recorded as predominantly male in one country and predominantly female in another, with at least 1,000 birth registrations and at least 80% one gender on each side. We checked 18,076 names, every name given to at least 500 babies in the United States since 1880, against the counted registration data of seven national statistics offices.

The largest case is Jean: 1,910,805 registrations in France, 100% male, against 38,983 in the United States, 95% female. Same four letters, opposite answers, both from official records.

The short version

• 47 of 18,076 names change gender between countries under strict thresholds: at least 1,000 registrations and at least 80% one gender on both sides.
• 27 of the 47 are male in France. 31 are female in the United States. The most common pair is male in France and female in the United States, with 19 names.
• The list is a floor, not a ceiling. Only seven countries publish counted name statistics, so a name like Andrea, male in Italy and female in Germany, cannot appear here: neither country publishes counts.

The full list

Each side shows the country where the name is most often registered with that gender, the share of that gender there, and the number of registrations behind it. Names with a published profile link to it.

What the list is made of

A few patterns account for most of it.

French masculine forms that English reads as feminine. Jean, Alexis, Patrice, Lilian, Jocelyn, Vivian and Vianney are male in the French records and female in the English-speaking ones. Laurence and Daniele run the other way: female in France, male in the United States and Spain respectively.

Spanish, Catalan and Galician male names. Joan (Catalan for John), Angel, Lois (the Galician form of Luis) and Toni are male in Spain and female in the United States or England and Wales.

Nordic names. Kim, Ola, Jo and Remi are male in Norway. Eli goes the other way: female in Norway, male in Canada, England and Wales, the United States and France.

Arabic and West African names. Reda, Rida and Alpha are male in France and Spain. In the United States the same spellings were given mostly to girls.

Surname-style names. Ashley, Henley, Bertie, Mckenzie, Bayley and Presley are female in the United States or Canada and male in England and Wales.

What this means if you infer gender from names

A country-blind lookup gives each of these names one answer, so it is wrong for one whole population. If 20% of a customer file is French, every Jean in that segment is misclassified, while the error rate for the file as a whole still looks acceptable.

Passing the country fixes it for every name in the table:

curl "https://namegender.com/api?name=Jean&country=FR" -H "Authorization: Bearer YOUR_KEY"
# gender: male, probability: 100, total_names: 1910805

curl "https://namegender.com/api?name=Jean&country=US" -H "Authorization: Bearer YOUR_KEY"
# gender: female, probability: 95, total_names: 38983

For a spreadsheet, the same thing is a country column in the upload; see adding a gender column to a spreadsheet. For a name that appears in several countries, the /countries endpoint returns every counted country with its own gender and share; the API reference has the details.

How the list was built

1. Candidates. Every name with at least 500 registrations in the US Social Security Administration files from 1880 to 2024: 18,076 names. Starting from US records means names that are rare in the US are missing, even if they change gender elsewhere.
2. Counts. For each candidate, the counted registrations in the seven countries that publish them: the United States (SSA, 1880 to 2024), England and Wales (ONS), France (INSEE, 1900 to 2025), Canada (StatCan), Spain (INE), Ireland (CSO) and Norway (SSB). The periods differ by country, so a share describes that country's whole published record, not a single year.
3. Threshold. A name is listed only if at least one country records it as male and at least one other as female, each with at least 1,000 registrations and at least 80% one gender. Loosening either threshold makes the list much longer and much noisier.
4. Date. Computed on 11 September 2026 from dataset version 2026.08, the same data the API serves.

A registration count describes how a name was recorded, not how any individual identifies. The list is useful for one purpose: showing where a single worldwide answer is guaranteed to be wrong for a large group of people.

Related: why the country parameter exists and our coverage by writing system, including where it fails.