← Back to Mini tools

Names, with a basis.

These are Census-weighted fictional names: a practical way to give classroom hypotheticals a broader cast, using public national statistics instead of a handpicked name list. They are not a statistically representative sample of real people.

The sources

The U.S. Census Bureau released Frequently Occurring First Names and Last Names in the 2020 Census on April 14, 2026. We use its nonnegative-count tables for first names by sex, first names by race and Hispanic origin, and surnames by race and Hispanic origin. These are aggregate tables, not individual records.

Population weights come separately from Table 4 of the Census Bureau’s 2020 supplementary tables on race and Hispanic origin. Name counts should not themselves be used to estimate demographic population shares.

How a name is drawn

  1. Choose a population category using the 2020 national population counts, or use the category you selected.
  2. Choose a given name in proportion to its published frequency within that category.
  3. If you select female- or male-associated names, multiply each name’s category frequency by its female or male share in the separate sex table.
  4. Choose a surname independently, weighted by frequency within the same population category.
  5. Reject identical full-name duplicates within your cast.

The national mix assigns Hispanic/Latino origin a probability of approximately 18.73% for each draw. Small casts can vary widely; the tool does not impose quotas or promise a particular mix in any one batch.

Population weights (2020 Census)
CategoryPeople
White alone, non-Hispanic191,697,647
Black alone, non-Hispanic39,940,338
American Indian / Alaska Native alone, non-Hispanic2,251,699
Asian / Pacific Islander alone, non-Hispanic20,240,737
Two or more races, non-Hispanic13,548,983
Hispanic / Latino, any race62,080,044
Some Other Race alone, non-Hispanic1,689,833

Asian and Native Hawaiian / Other Pacific Islander population counts are combined to match the available name table. The name tables have no separate non-Hispanic Some Other Race column. For that approximately 0.51% of national draws, the tool uses pooled name frequencies.

What the model cannot tell us

The Census publishes separate first-name-by-sex and first-name-by-background tables, not the joint distribution of all three. The sex filter therefore makes an approximation: sex and background are treated as independent given a first name. Population category weights do not change with this filter. Female and male refer to associations in the Census sex table, not a person’s gender identity.

Given names and surnames are published separately. Pairing them is synthetic and can sometimes produce culturally unexpected combinations. A name does not establish anyone’s race, ethnicity, nationality, sex, or gender. This tool is for inventing characters, not classifying real people. Coincidental matches to real names are possible.

Coverage and limits

For fast loading, the bundled data retain 19,784 given names and 12,606 surnames: the union of frequent names across each population and sex-associated pool. This covers 96.60% of individually published given-name counts and 73.20% of individually published surname counts. Including rare-name aggregates, coverage is 90.59% and 64.32%, respectively. These are shares of the source name counts, not shares of all U.S. residents.

The underlying release excludes names that did not occur at least 100 times in both the 2010 and 2020 Census. Missing responses, Census edits, privacy noise, and the common-name subset all limit coverage. Asian / Pacific Islander given-name coverage, for example, is 92.35% of individually published counts, but 70.52% when rare-name aggregates are included. Spelling is converted from uppercase to title case; missing accents are not reconstructed.

See the Census Bureau’s given-name methodology and surname methodology, plus this build’s coverage report and source checksums.

Local, fast, and inspectable

Both tools run entirely in your browser. The data are bundled with the site; generating names makes no network request. Student lists are held only in tab memory, with no analytics or persistence. Random group assignment uses a Fisher–Yates shuffle with browser cryptographic randomness and rejection sampling to avoid modulo bias. Groups differ in size by at most one. Duplicate roster entries are retained and flagged.

The source is directly inspectable: grouping and parsing, name sampling, and bundled aggregate data. Data build: November 17, 2026.

← Back to the tools