ONS Baby Name Tables: Spaces, Hyphens, Accents and Missing Rows

Keep the personal name and the statistical row separate

You may be looking for the spelling your family intends to use, while the statistical table publishes a processed form. Write the intended name exactly, including its spaces, accents and hyphen. Keep a second field for the row you find. Changing a search key does not change the personal name.

The ONS QMI says its baby-name statistics are based on the first names on birth certificates and do not combine names merely because they sound similar. The same report documents cleaning before tabulation. “Exact spelling” therefore needs to be read together with those cleaning notes, not as a promise that every character in the original record appears unchanged in the table.

This guide explains the published methodology reviewed here. It does not approve a name for registration, inspect your birth certificate or implement ONS’s internal software. For a document-entry question, use the responsible authority’s instructions.

Read hyphenated names on the correct side of 2018

In its accuracy section, the QMI explains that the earlier processing removed spaces and the text following a space. From 2018, it removed the spaces in hyphenated names instead, keeping the hyphenated form together. ONS gives Amelia - Lily → Amelia-Lily as the example.

That example is useful when a compound name seems to appear in one period and disappear in another. Do not automatically interpret the change as a change in parents’ choices. First ask whether the row definition changed. ONS says the revised cleaning slightly changed the number of unique names in 2018; a methodological discontinuity can affect a series even when the underlying personal spelling is unchanged.

The report does not supply a complete character-by-character transformation table. It does not establish here how every nonhyphenated multiword name, apostrophe, dash character or unusual space is treated in every edition. If your question depends on one of those details, read the edition notes or seek clarification rather than extending the example into a universal parsing algorithm.

Accent removal is a processing rule, not a meaning claim

The QMI states that accents are removed during automated processing because its software does not recognise them. This is a statement about that statistical pipeline. It does not mean accents are optional in a person’s written name or that changing an accent preserves pronunciation and history.

Do not assume the report specifies a particular Unicode normalization procedure or every replacement character. A generic accent-removal script would be your own implementation, not a verified copy of ONS’s processing. Keep the source wording and the matched row visible in your notes.

W3C’s personal-name guidance gives a different instruction for forms and databases: allow the name’s actual spelling and avoid restrictive assumptions about spaces, punctuation, case and scripts. A website that collects a personal name should not automatically adopt a statistical-cleaning rule as its display spelling.

If two written forms look similar, research their histories and intended pronunciations separately. A statistical match cannot establish those facts. Use the language-origin guide when that is the actual question.

Use a five-step row check

1. Choose the birth year and recorded category. Start from the official boys’ or girls’ dataset page. The reviewed pages list a 2025 edition and earlier downloads; an edition year is different from its release date.
2. Read the notes and table title. Confirm whether you are looking at all published names, the top 100, a country table or another breakdown. A name missing from a top-100 table may still have a published row elsewhere.
3. Search the intended spelling first. Record the exact typed form. If the documented processing is relevant, try the described row form as a separate search; keep both in the record.
4. Copy the matched row faithfully. Keep the name label, category, geographic scope, count, rank and edition together. Do not combine a count from one table with a rank from another.
5. Leave unresolved cases unresolved. Record the checked tables and relevant limits rather than converting a failed search into a count of zero.

This is an original reader workflow. The workbook interface and individual rows were not tested for this article. A search function may use case or accent rules of its own, so distinguish the matching behavior of your software from the source’s published row.

A missing row is not a count of zero

The reviewed QMI says ONS withholds individual names with fewer than three occurrences in England and Wales as a whole. A name absent from a public all-name table can therefore have a small nonzero count. A narrower table can omit names for additional reasons, such as publishing only the top 100.

The report also describes [x] in rank comparisons as a marker for a name not previously included because its count was below three. That is not a number to enter as zero. Read the workbook’s own legend before interpreting a marker in a particular edition.

If an accented or compound spelling is absent, several questions remain: did cleaning change the row key, is the chosen table exhaustive for published names, and did disclosure prevent publication? Those possibilities cannot be separated from a missing search result alone.

When comparing a series across 2018, keep a methodological note beside the data. Avoid adding a pre-2018 simple-name row to a later hyphenated row and calling the result one consistent exact-name history. You need a defensible definition and sufficient data before making that comparison.

Work through four constructed search situations

These situations are invented reader cases. Only the Amelia - Lily transformation itself is the example supplied by ONS.

Keep a short source record

Personal spelling exactly as intended:
Birth year, category and geographic table:
Edition and source URL:
Table scope and relevant legend:
Search spellings tried:
Published row label found:
Count and rank copied from that row:
Processing note or 2018 boundary:
Unresolved disclosure or character question:

The result can be “matched this processed row,” “not in this top-100 table,” or “unresolved after these checks.” Those statements preserve what you actually learned.

For the broader question of comparing countries, continue to the U.S. and England–Wales comparison guide. For a family’s initials, use the complete-name checklist, retaining the actual personal spelling.

Download the original UTF-8 record and keep your own completed copy.