Costs
Count, share or rank: what a naming trends timeline plots
A naming trends timeline can plot counts, shares or ranks, and each answers a different question. Here is how to pick one and word the claim.
A naming trends timeline can plot three different quantities from the same annual tables: how many babies got a name, what share of births that was, or where the name ranked. Each answers a different question, and a chart that does not say which one it plots cannot be checked.
For the United States, the annual tables almost always come from one publisher. The Social Security Administration's national baby names data is published as Baby Names from Social Security Card Applications - National Data. The choice of quantity is then yours, or the chart maker's.
What to take away
- A count follows the number of births, so it can fall while a name holds its share.
- A share is the quantity that matches a claim about preference.
- A rank is a position in a list and says nothing about how many babies carried the name.
- A smoothed line can turn a one-year step into a slope.
Three quantities, three questions
Pick the quantity from the claim, not from the chart you happen to have. A claim about how many babies were given a name needs counts. A claim about preference needs shares. A claim about standing among other names needs ranks. The table sets out what each can carry.
| Quantity | The question it answers | Where it misleads |
|---|---|---|
| Count | How many babies were recorded with the name | Falls when the birth cohort shrinks, even if preference holds |
| Share | What portion of recorded births carried the name | Needs the matching total for the same year and the same series |
| Rank | Where the name sat against the others that year | Hides the size of the gap between neighbours |
| Smoothed value | The broad direction over several years | Spreads a one-year step across the window |
Label the axis with the quantity, and keep the label when the chart is cropped or reposted. A chart copied without its axis label is how a count turns into a share by the time it reaches a second article. If the claim covers announcement years, sort the announcements by year first, so the dates and the series cover the same window.
Why share fits a claim about preference
A share fits a preference claim because it divides by the cohort. Cohort size changes year to year for reasons unrelated to naming. When it falls, every count falls, and a name holding exactly its share of a smaller cohort looks as if it is declining.
The reverse also holds. When the cohort rises, ordinary names look newly popular. A count going down while a share goes up is not a contradiction; it is what happens when the population shrank. Say which of the two you are quoting, and keep them out of one paragraph.
The same question belongs on every story about a name that surged after a famous family chose it. Ask whether the surge is a count or a share. The page on celebrity baby names by year covers what national tables can and cannot show about such claims.
Why a rank says so little
A rank is a position, not a measure. Ranks in the middle of a long tail move a great deal on tiny frequency differences, and ties are broken by an arbitrary rule. Read a rank as a place in a list.
Rank also hides the gap between two lists. The ONS baby names bulletin for England and Wales states that the top names for boys had much higher counts than those for girls. A rank of one on each list therefore stands for different counts, and the rank alone cannot say by how much.
A name that jumps several places may have gained only a few babies if its neighbours sit close together. Put the counts beside the ranks before calling it a rise, and report the counts when the ranks are all you were given.
A step is a question for the agency
Smoothing makes a chart readable and removes the feature that matters most here, which is the discontinuity. A multi-year average spreads a step across its window, so a sudden change reads as a slope. Look at the unsmoothed annual figures first, and smooth only for presentation.
Is a sharp move real?
Does the series step in one year?
Check what changed at the agency
Gradual movement, naming behavior
Naming changes gradually, because it comes from millions of independent decisions, and those do not synchronize. For any sharp movement, ask what changed at the agency, not in the culture. The unsmoothed line shows where the step sits, and the agency's notes say whether the counting changed that year.
A coincidence in time is not a cause. When the event is a famous birth, the birth announcement timeline supplies a date to test against. A timeline can show that a change followed the event, but timing alone cannot establish that the event caused it.
Where a stitched series can break
A timeline is a stack of annual tables joined by whoever drew the chart, and each join assumes this year's table measures what last year's did. The figure sets out four joins and the assumption behind each, from year to year through to published and provisional editions.
Where series joins break
Join
- Year to year
- Counting rule unchanged
- Geography
- Area means same thing
- Variant
- Same spellings grouped
- Published
- Editions comparable
Assumption
- Year to year
- Rules revised, documented not announced
- Geography
- Boundaries redrawn between editions
- Variant
- Grouping differs by publisher, year
- Published
- Provisional and final figures differ
How it fails
- Year to year
- Geography
- Variant
- Published
Plotting a share does not repair a broken join. If the counting rule changed between two editions, the count and the total changed with it. The step then belongs to the instrument, and the unsmoothed annual figures are where it shows.
Registration figures count registered births, while an announcement is a different record with its own date, as birth announcements facts explains. Keep the two apart when a chart is set against announcement dates, and say which one each date comes from.
Writing the claim in five steps
A defensible time claim is a duller sentence than most coverage carries, and nobody has to correct it later. Build it in this order and stop at the fifth step.
- Name the series and its publisher.
- State the quantitycount, share or rank.
- Give the years and say whether any are provisional.
- Say which comparability breaks fall inside the window.
- Stop short of a cause unless you have one.
Such a series can describe patterns across a population, but it cannot explain why a particular family chose a name. It can show that a name rose in the same years as an announcement, and never that one family moved it. Check the announced spelling against the family's own statement, as celebrity baby names announcements describes.
Common questions
Should I plot counts or shares over a long comparison?
Shares, for any claim about preference, because the cohort behind the counts changes over the years. Treat even that as a broad shape. Recording practice and the population counted changed enormously, so a long comparison measures the record-keeping as much as the names.
Can a ranked list from a news article stand in for the table?
Only as a pointer to the table. A rank gives position and omits the counts, so go back to the agency table and say so if you could not. A back-series assembled from write-ups inherits every definition that got dropped along the way.
How many years do I need before I can call something a trend?
Enough to see the movement continue through at least one revision cycle, which in practice means several. Two points make a line, and they tell you nothing about whether the line keeps going. Name the quantity and the years, and let the reader judge the length.
Can a timeline show that a famous family moved a name?
No. It can show a name rising in the same years, and nothing more. For the wider picture of names, announcements and what the records confirm, start from the guide to celebrity baby names and check each claim against its own source.
