About the data: United States sources, method and limits

Every list on this site is computed from public records, not from other websites. This page says which records, how they are used and where they fall short.

Sources

Licenses are recorded as stated by the publisher. Check the publisher pages above before reusing the raw files.

How a name is picked

• First names come from Social Security card applications for babies born in the United States. The file holds 117,791 distinct name and sex combinations over 375,311,072 recorded births, 1880 to 2025. A first name is drawn with probability proportional to its number of births in the period of the page.
• A last name is drawn independently from the 2000 Census surname list, in proportion to the number of people who carry it. First and last name do not influence each other, so the tool does not know that some combinations are more likely than others.
• The Realism slider raises each weight to a power. At 1 the picks follow real frequencies, at 0 every name is equally likely.
• All random choices come from one seed. The same seed, options and page always produce the same list, which is why the link under the generator reproduces a list exactly.

How much of the data each list covers

• First name lists keep the most common names until they cover 95% of recorded births for each sex (85% at the lowest if the file would be too large). Rarer names are left out to keep each page's download small.
• The surname list keeps the 10,797 most common of 151,671 surnames. That is 77% of the surname counts in the Census file, which itself covers 89.8% of the population. A full 95% would take a download about five times larger than we allow.

What the data cannot tell you

• The Social Security file leaves out names given to fewer than 5 babies of one sex in one year, so very rare names are missing. It also holds placeholder entries for unnamed babies, such as Unknown or Baby; those are removed before any list is built.
• The file covers people who applied for a Social Security number, mostly people born in the United States. It counts births, not living people, so the list for 1940 to 2025 is not adjusted for who is still alive or who moved.
• Surnames are from 2000, not from the same years as the first names. Surname patterns have shifted since then, for example towards more Hispanic and Asian surnames. The 2010 Census top 1,000 surnames agree closely with the 2000 ranking (Spearman rank correlation 0.963), but the 2010 file is used only as a check, not as a source of picks.
• Regional differences, name and surname pairs that go together, and middle names as a separate statistic are not in the data yet. A middle name, when you ask for one, is another draw from the first name list.

Pages that are not built

A page is built only if its data passes a check: enough names, a recorded source, and text that differs enough from sibling pages. Every planned page passed in the latest build.