Data sources
UK figures are the site’s records from the Office for National Statistics baby-name data for births registered in England and Wales. US figures are the site’s records imported from the Social Security Administration national baby-name files.
Exact spellings are analysed separately. The verified production table is baby_name_yearly_stats, filtered by country_code and gender. No Trends Laboratory report writes to that table.
How the latest year is detected
The code groups existing rows by country, gender and year, then uses the latest year containing a usable rank or birth count. It does not use the current calendar year and does not rely on a year embedded in the URL.
“Latest five years” means the five latest annual datasets present for that country and gender. A one-year report is the exception: it compares the latest annual dataset with the immediately preceding available dataset.
Rank and birth-count movement
For a rise, places gained equals the starting rank minus the latest rank. For a fall, places lost equals the latest rank minus the starting rank. Lower rank numbers indicate greater popularity.
Absolute birth change is latest births minus starting births. Percentage change is that difference divided by starting births. If the starting count is zero or unavailable, no percentage is calculated.
Missing, redacted and newly recorded values
ONS source values of two or fewer may be redacted. The database stores those records with is_redacted = 1 and null rank/count values. The SSA import code stores published rows with is_redacted = 0. In either dataset, an absent record is not assumed to mean zero births.
A name present at the end but without a usable starting record is listed separately. If it appeared before the starting year it is described as returning; otherwise it is described as newly recorded. Names missing at the end are also separated rather than assigned a fabricated final rank.
Comeback detection
A comeback must have at least 8 ranked years, a historical fall of at least 100 places, a recovery of at least 50 places, at least 20 births in the latest dataset, and 2 consecutive recent rank improvements.
The former peak is chosen before the recent recovery window. The low point is the least popular rank after that peak and before the sustained recent rise.
Volatility score
Only consecutive-year rank changes are used. The score is the mean absolute annual rank movement plus half the standard deviation of those annual movements. Results require at least 8 ranked years, 75% continuity and 20 latest-year births.
UK and US alignment
Direct rank comparisons use the latest year that exists in both datasets for the selected gender. Recent direction is calculated independently over each country’s latest 3 annual datasets, and both periods are printed on the page.
A lead in one dataset may be worth watching, but it is not presented as a prediction. Registration systems, populations and publication rules differ, so raw UK and US birth totals are shown for context rather than treated as directly equivalent market shares.
Central data-quality thresholds
- Minimum relevant birth count: 20
- Minimum rank movement: 10 places
- Minimum percentage movement: 5.0%
- Maximum rows displayed per principal table: 50
- Minimum years for long-history analysis: 8
- Minimum decade coverage: 4 annual datasets
These rules reduce distortion from very small or fragmentary series. They also mean a report is not an exhaustive list of every change in the source table.
Why results change after an import
Report URLs are timeless. Their calculated data is cached without an automatic expiry so repeat visits avoid the expensive database work. After importing a new annual dataset, delete the generated JSON files in /trends/cache/; the next visit to each URL rebuilds it from the latest database records.
Trends describe recorded patterns; they do not prove why parents chose a name and they do not guarantee what will happen next.