What's in a Name? / Toban Wiebe

How it works

The Social Security Administration publishes national baby-name counts by first name, sex, and birth year. For a selected name and sex, this app filters the data to those records and normalizes the counts across birth years.

The calculation is done separately after choosing female or male. Holding sex fixed without loss of generality, write $Y$ for birth year, $N$ for first name, and $C n,y$ for the SSA count for name $n$ and year $y$ . Bayes rule gives:

\begin{aligned} P(Y = y \mid N = n) &= \frac{P(N = n \mid Y = y)P(Y = y)} {\sum_t P(N = n \mid Y = t)P(Y = t)} \\ &= \frac{P(N = n, Y = y)} {\sum_t P(N = n, Y = t)} \\ &\approx \frac{C_{n,y} / C_{\cdot,\cdot}} {\sum_t C_{n,t} / C_{\cdot,\cdot}} \\ &= \frac{C_{n,y}}{\sum_t C_{n,t}}. \end{aligned}

The joint probabilities are estimated by the corresponding SSA name-year count divided by the total count in the fixed-sex data. That common denominator cancels, so the app only needs the selected name's counts by year.

The chart is therefore a birth-cohort distribution. It is not adjusted for mortality, immigration, or emigration, so it should not be read as the exact age distribution of living people with that name today.

The popularity chart uses a birth-weighted percentile. In a given year, a percentile of 75% means that 75% of babies in the selected sex group had names that were as rare or rarer, while 25% had more common names. This is weighted by births rather than by distinct name labels, so very rare names do not dominate the scale.

SSA suppresses low-count records for privacy. Names with fewer than five births in a year/sex cell can be missing, and very rare names may have no usable data.

Source: Baby Names from Social Security Card Applications - National Data .

What's in a Name?

Estimate a birth cohort

How it works

Keyboard