Naming trends source guide: SSA and ONS data

Naming data is unusual among the datasets a general writer touches, because the agencies that publish it document their own limitations carefully and almost nobody reads the documentation. The table gets used. The notes beside it, which say what the table counts and what it silently excludes, do not.

This naming trends source guide covers two national series by name: the Social Security Administration baby names file for the United States and the Office for National Statistics baby names release for England and Wales. For each it explains what to read before the numbers, what the series actually covers, and what to record so a figure you publish can be traced back years later.

What to take away

Read the construction notes first

Take the American series as the worked example, because its documentation is unusually plain. That series is the baby names file of the Social Security Administration, the United States agency that publishes national name counts.

The Social Security Administration's published background on how the name data is built states where the records come from, which cases are excluded, and what the tabulation rules are. Several of those rules will surprise a writer who has only seen the rankings.

Spellings are counted separately, so variants of one name become separate entries, not a single popular name. Spaces and hyphens are removed before counting, merging forms a family considers distinct.

Names are tabulated by sex as recorded and are not edited, so entries that are clearly not names stay in the data. Ties break alphabetically, so a rank difference can mean no difference in frequency.

None of that is a flaw. It is a set of decisions, published, and a writer who has not read them will make claims the data does not support. The spelling-variant rule matters most: a "rising" name may simply be one spelling gaining on another.

What each source actually covers

The first row is the Social Security Administration file. The second is the Office for National Statistics release for England and Wales, which uses birth registration data. For the third row, the same ONS release breaks names down below the national level: the ONS describes its product as the release that provides the most popular names by area, counted by the unitary authority of the mother's residence.

The bottom row is the trap. An international series makes countries comparable by redefining what is counted, and a reader who assumes the national definition will draw a conclusion that neither source supports.

The World Health Organization's framework material on civil registration and vital statistics systems is worth reading precisely because it shows how much national variation there is under a single label.

Any bulletin a statistical office publishes carries its own equivalent notes.

The Office for National Statistics release on baby names in England and Wales states its coverage, reference period, and treatment of variants. Those differ from the Social Security Administration rules enough that a direct comparison between the two series is wrong, not merely rough.

Reading a trend in the ONS release

The 2023 edition shows what a trend statement in this series looks like. The ONS reported that the three most popular names for baby girls in England and Wales were Olivia, Amelia and Isla, unchanged since 2022, and that Muhammad overtook Noah as the top name for baby boys, followed by Noah and Oliver.

Read that as a statement about registered births in England and Wales, where the ONS counts unique names and spellings. It says nothing about the United States, and another spelling of the same name is a separate entry.

Note too that the ONS has discontinued the written bulletin. It continues to publish the data in its Baby names for girls and Baby names for boys datasets, so the dataset edition, not the bulletin, is what a citation should name from now on.

Suppression, and the names you will never see

Every serious name series withholds low-frequency entries, because a rare name in a small area identifies a specific child.

The Social Security Administration files state the rule directly: entries with fewer than five occurrences in a geographic area are excluded so no individual can be picked out. The notes on what lies beyond the published top names set this out.

The ONS applies the same principle to its area tables. Names with low counts are suppressed to protect the identity of an individual, consistent with Section 39 of the Statistics and Registration Service Act 2007, which makes it a criminal offence for the ONS to disclose information that would identify an individual.

The practical consequences are worth stating because writers trip over all three.

State counts will not sum to the national count, and the gap is the suppressed material rather than an error. An unusual name may be entirely absent, which is not evidence that nobody used it. And a small-area analysis of rare names cannot be done from published data at all, by design.

The argument in full belongs to the question of what a profile may carry.

Recording a source so a number can be traced

A figure on a page outlives the tab it came from. Six fields make it recoverable.

The six fields:

The last field prevents most of the errors. A rank and a count answer different questions, and a page that stores only the number loses which one it had.

Worked example. Social Security Administration, Baby Names, national file, 2023 edition. Population: births recorded in the United States and filed with the agency. Definition: count of occurrences that year, with spellings counted separately. Figure: the count printed on the name's row, copied from the edition rather than rounded or estimated.

The same six fields, filled for England and Wales: Office for National Statistics, Baby names for girls in England and Wales, 2023 edition. Population: births registered in England and Wales. Definition: exact spellings, with low counts suppressed. Figure: the rank or count printed on the row.

The same discipline applies to any source, including the family references we cite.

The secondary source that lost the definition

Most naming coverage cites another article rather than a dataset. The definition is what goes missing, and it goes missing at every hop.

An agency publishes a table with notes. A news outlet drops the notes. A second outlet drops the geography. A third rounds the figure. By the fourth hop a bare number sits beside a claim the original notes ruled out.

The fix: reach the release itself. These bodies publish openly, so it is usually quick. If you cannot, say you are citing an intermediary and name it. That is the standard for any chain source, and it is the same rule that applies when tracing sibling claims to their earliest version.

Whatever the series, a count describes a population of births; it does not establish anything about one named child.

Common questions

Which source should we treat as authoritative?

The one whose population matches your claim. There is no globally best series; there is a series that covers the jurisdiction and cohort you are writing about, and using any other is a category error however reputable it is.

The agency reissued last year's table with different numbers. Which is right?

Both, for their editions. Provisional figures are replaced by revised ones as registrations complete, and citing the edition alongside the figure is what stops a revision from making you look wrong.

Can we combine two countries' tables into one ranking?

Almost never usefully. The counting rules differ, the variant handling differs, and the suppression thresholds differ, so a combined ranking measures the differences between the systems as much as anything about names.

How current is current?

Later than most coverage implies. Registration, coding, and publication each take time, so a series describing a birth year appears well after that year closes. Any claim about the present is an extrapolation and should be written as one.