Counts with counts
National published counts are compared only across 2000, 2010, and 2020. The 2020 total is derived from published, noise-affected components.
Methodology
How to compare the four surname files
Each Census vintage answers a slightly different question. This site preserves those differences instead of forcing every year into one continuous measure.
Comparison rules
1990 stays separate
The 1990 product is sample-based and supplies frequency percent and rank, but not a national surname count.
Absence is not zero
A surname missing from a published file is labeled unpublished or ineligible. It may still have existed in the population.
Vintage-specific methods
1990
Sample frequency
Population universe: Valid surname records in the 1990 Post-Enumeration Survey Search Area sample, not the full decennial census population.
Publication rule: The downloadable surname file contains 88,799 edited sample surnames and reports sample frequency through the available distribution; absence does not mean nonexistence.
Count semantics: not provided
2000
Published surname table
Population universe: People with valid surname information in Census 2000 after documented edits.
Publication rule: The public all-name file contains 151,671 surnames occurring at least 100 times; demographic cells from 1 through 4 are suppressed as (S).
Count semantics: published source count
2010
Published surname table
Population universe: People with surnames recorded in the 2010 Census after documented edits.
Publication rule: The public table contains 162,253 surnames with frequency at least 100; the ALL OTHER NAMES aggregate is not a surname entity.
Count semantics: published source count
2020
Disclosure-protected published cohort
Population universe: Edited 2020 Census last names that were also in the 2010 public names table and had an unprotected frequency of at least 100 in both years.
Publication rule: The public workbook contains 156,621 cohort-eligible names. Disclosure-protection noise is added to race and Hispanic-origin component counts; total count and proportions are calculated from those noisy components, so a few published totals fall below 100. Public-file reconciliation finds 156,620 of those names in the downloadable 2010 CSV; NA is the sole exception and is flagged rather than zero-filled or silently removed because the methodology's eligibility process can rely on the underlying 2010 names table rather than this public-file intersection.
Count semantics: derived from published components