The Census · Directive 15 · The Naturalization Cases · Virginia · South Africa · France
Race and Ethnicity: What the Record Shows About the Government's Classification of Americans
The federal government has classified Americans by race since the first census in 1790, and the classifications have never held still. Americans of Middle Eastern and North African origin are White under the standard issued in 1977 and kept in 1997; a revision published March 28, 2024 moves them into a category of their own beginning with the 2030 census; a rule proposed September 10, 2026 would strike the race and ethnicity questions from that census entirely. The courts once decided whiteness applicant by applicant, Virginia's registrar reclassified a people's records by surname, and every regime that held the pencil described its own drawing as progress.
The United States government sorts its residents into racial and ethnic categories of its own design, and it revises the design on its own schedule. Under the federal standard issued in 1977 and reaffirmed in 1997, a person having origins in the Middle East or North Africa is White, and has been White in federal statistics for nearly fifty years; more than 3.5 million people reported Middle Eastern or North African descent inside the White category in the 2020 census. A revision of the standard, published in the Federal Register on March 28, 2024, redraws the map: seven minimum categories where the old standard kept five races and a separate ethnicity, one combined question where the old standard asked two, Middle Eastern or North African standing as a category of its own, detailed origins collected by default, full compliance due by 2029, and the 2030 census as the first count under the new design. Implementation has since slipped twice, the Census Bureau has withdrawn its plan to introduce the new questions on its largest survey in 2027, and a rule proposed on September 10, 2026 would bar race and ethnicity questions from the decennial census altogether. The people being classified account for none of these changes. The record behind the current filings runs through two centuries of the same operation: a naturalization statute that admitted “free white persons” and left the courts to decide who they were, applicant by applicant; a federal dictionary of 1911 that counted South Italians as a race apart from North Italians; a Virginia registrar who reclassified his state's Indian families as colored by circular and surname list; a one-drop rule written into a state statute and into census instructions; and, abroad, a South African classification office that had reclassified 18,469 people by 1956 and a French republic whose law forbids the questions entirely. Every version of the system arrived carrying a justification in its era's moral vocabulary. The justifications are part of the record, and this report prints them.
The map in force
The standard that governs most federal data collection today is Statistical Policy Directive No. 15, first issued by the Office of Management and Budget in 1977 and revised once, in 1997. It sets five minimum race categories and treats Hispanic or Latino origin as an ethnicity asked in a separate question. Its definition of the largest category reads, in full: “White. A person having origins in any of the original peoples of Europe, the Middle East, or North Africa.” The Census Bureau's worked examples under that definition list, side by side, entries such as Irish, German, Italian, Lebanese, Near Easterner, Arab, and Polish. The standard states its own nature on its face: the categories, in its words, “are social-political constructs and should not be interpreted as being scientific or anthropological in nature,” and its stated field of use is “Federal statistics, program administrative reporting, and civil rights compliance reporting.” For most of the census's history the answer to the race question was supplied by an enumerator's observation; self-identification spread with the mail-out census around 1970, and the 1977 directive made the respondent the source of the answer as a matter of federal principle. The 1997 revision added one further liberty: beginning with the 2000 census, a respondent could mark more than one race.
The outputs of that map, as measured in 2020, show the instrument at work. The Bureau counted 204.3 million people as White alone and 235.4 million as White alone or in combination. The residual category, Some Other Race, counted 49.9 million alone or in combination and passed the Black or African American population, 46.9 million, as the second-largest group in the country. The population marking two or more races rose from 9 million in 2010 to 33.8 million in 2020, an increase of 276 percent, and the Bureau attributes a large share of the movement to its own redesigned questions and coding procedures, cautioning that comparisons across the decade “should be made with caution” for exactly that reason. The Bureau's linked-record research adds the individual-level figure: matching the same people's answers across the 2000 and 2010 censuses, Liebler and her co-authors found that about 9.8 million people, 6.1 percent of 162 million linked records, gave a different race or Hispanic-origin answer in 2010 than they had given in 2000. The categories held still for that decade. The people inside them moved anyway.
The revision of March 28, 2024
The notice published March 29, 2024, at 89 Federal Register 22182, announced revisions decided the day before, the first since 1997. The revised standard carries seven minimum categories: American Indian or Alaska Native, Asian, Black or African American, Hispanic or Latino, Middle Eastern or North African, Native Hawaiian or Pacific Islander, and White. Race and ethnicity merge into a single question, with respondents directed to select all that apply. The definition of White contracts to origins in Europe. The new category is defined by origins in the Middle East or North Africa, and the standard requires collection of detail beyond the minimum categories by default: printed example groups and write-in lines under each heading, the examples drawn from the largest groups measured in 2020—Lebanese, Iranian, Egyptian, Syrian, Iraqi, and Israeli under the new heading; English, German, Irish, Italian, Polish, and Scottish under White. The notice states the purpose in its summary: the revisions “are intended to result in more accurate and useful race and ethnicity data across the Federal government.” Agencies were directed to publish implementation plans within eighteen months and to bring every existing collection into compliance by March 28, 2029. The Census Bureau's implementation schedule made the 2030 census the first decennial count under the new map, with the new questions entering the American Community Survey ahead of it.
The case for the new category, as its advocates presented it across more than three decades of campaigning and in the revision's public comment file, is a case about standing in the data: a population counted inside another category cannot be measured for civil-rights enforcement, health disparities, or eligibility for programs allocated by the numbers. Reporting on the announcement recorded the campaign's length—advocates for Arab American and other Middle Eastern and North African communities had sought a category of their own since the standards were young—and the Census Bureau's own research supplied part of their evidence. In the Bureau's American Community Survey records, one in four people of Middle Eastern or North African ancestry did not report White alone, although White was the answer the governing definition assigned them. The 2024 notice's testing record shows what happened when a category was offered: in the Bureau's 2015 National Content Test, respondents of these origins used it in large numbers. The same footnoted record shows the new box fitting some of its assigned peoples loosely. Armenian respondents in panels offered the category chose White 79.0 percent of the time and Middle Eastern or North African 12.6 percent; Sudanese respondents chose Black or African American 87.2 percent and the new category 8.0 percent. The pencil drew a region. The region's peoples, in the government's own test data, distribute themselves across three of its boxes.
The pencil moves again, 2025–2026
The 2024 map stands as the governing standard, and its arrival has been postponed twice. Implementation trackers maintained by the new category's advocates record that the Office of Management and Budget first moved the agency action-plan deadline from September 2025 to March 2026, then to March 2027, and moved the government-wide compliance date from March 28, 2029 to September 28, 2029, seven months before the 2030 census. The Census Bureau's survey schedule moved with it: a December 2025 Federal Register notice proposed adopting the new questions in the 2027 American Community Survey; the supporting documents filed with a May 14, 2026 notice state that the Bureau will not implement the standards in 2027. Then, on September 10, 2026, the Department of Commerce and the Census Bureau proposed a rule, 91 Federal Register 57524, that reaches past delay to removal. Its proposed Part 60 rewrites the residence criteria for the apportionment count. Its proposed Part 70 provides, in the text quoted across the comment file: “no questions about race, ethnicity, or sexual orientation shall appear on the short form of the decennial census questionnaire or any questionnaire used for the enumeration of the population.” The rule's summary defends the removal on the ground that the count should not be, in its phrase, distorted by questions about “immaterial personal characteristics, such as race,” and its preamble asserts that the sole constitutional purpose of the decennial census is the apportionment of the House of Representatives.
The government's own filings argue with the proposal. The Bureau's preliminary regulatory impact analysis, quoted in the comment record by the National Women's Law Center and by a letter from twenty-four United States senators, states that fertility, mortality, cancer, poverty, unemployment, and labor-force rates are produced from decennial counts by race and ethnicity because there are “no other comparably accurate, comprehensive, and reliable sources of this information,” and the senators' letter quotes the filing's own conclusion that the rule “deteriorates the accuracy” of those series. The 2024 notice, two years earlier, had justified the opposite course in the opposite vocabulary: the categories, OMB wrote then, are sociopolitical constructs, and they remain essential to accurate federal statistics and to evaluating how federal programs serve the public. The comment period on the 2026 rule, first set to close October 13, was extended to November 2, 2026 after tribal governments objected to the consultation schedule; the file as it closed carries opposition from the League of Women Voters and the National Network for Arab American Communities, the Native American Rights Fund, the National Academy of Education, and the National Women's Law Center, among others. As of this filing the rule stands proposed. No final rule appears in the record the desk reviewed. The desk therefore prints the state of the question exactly: one federal standard, revised in 2024 and unrepealed, orders a new category drawn for 2030; a second federal document, proposed in 2026 and undecided, would erase the questions the category would be asked in.
Whiteness, the fixed point
Every movement in this record is measured against a single category. The Naturalization Act of 1790 opened citizenship to “any alien, being a free white person,” and Congress extended the statute in 1870 to aliens of African nativity and persons of African descent, leaving every other people on earth to litigate its position relative to one word. Across the following eighty years the federal courts decided the question applicant by applicant, in dozens of reported prerequisite cases, and the two Supreme Court decisions that close the era show the test changing under the pressure of the cases. In Ozawa in November 1922, the Court equated white with “persons of what is popularly known as the Caucasian race” and held a Japanese applicant outside it. Three months later, in United States v. Bhagat Singh Thind, decided February 19, 1923, the applicant's ethnological evidence was conceded and then set aside: a high-caste Hindu of the Punjab, classified by the science of the day as Caucasian and Aryan, was held not white, in an opinion that states the rule in one sentence: “the words ‘free white persons’ are words of common speech, to be interpreted in accordance with the understanding of the common man, synonymous with the word ‘Caucasian’ only as that word is popularly understood.” The case itself was the government's suit to cancel a citizenship already granted. The science had admitted Thind. The common understanding removed him.
The traffic ran the other way for the peoples of the eastern Mediterranean. George Dow, a Syrian immigrant in South Carolina, was denied naturalization by a district judge who read the 1790 statute to cover persons of European nativity or descent and no others. On September 14, 1915, the Fourth Circuit reversed in Dow v. United States, holding on the “generally received opinion” that “the inhabitants of a portion of Asia, including Syria, were to be classed as white persons,” and giving the statute a construction broad enough to include “Syrians, Armenians, and Parsees” within the term. Dow is the direct ancestor of the line in the 1997 standard: the peoples the Fourth Circuit admitted to whiteness by opinion in 1915 are the peoples the Office of Management and Budget carried inside the White definition by directive in 1977, and the peoples the 2024 revision moves out of it. Absorption into the category also proceeded without litigation, by redefinition at scale. The Dictionary of Races or Peoples issued by the federal Immigration Commission in 1911 listed the peoples of Europe as separate races for official purposes and divided Italians at the Apennines: the North Italian “cool, deliberate, patient, practical,” the South Italian “excitable, impulsive, highly imaginative, impracticable,” two distinct ethnic groups in the Commission's ethnology. Federal immigration statistics of the era counted the South Italian as a race in his own right—1,911,933 South Italian admissions recorded between 1899 and 1910, of 2,284,601 Italian admissions in all. Within two generations the two races of 1911 sat inside the single White box of 1977, and the Census Bureau's definition examples print “Italian” with no compass point attached.
The direction of petition reversed across the century, and the reversal is documented at both ends. In 1915 the applicants sued for admission to the White category, because the category carried citizenship; the racial bar to naturalization stood until Congress removed it in 1952. In the decades before 2024, the organized petition ran outward: communities classified White by directive campaigned to be classified separately, because the category now carried invisibility in the data by which civil rights are enforced and disparities are measured. The desk records the constant beneath the reversal. Both petitions were filed with the authority that maintains the category. Both were argued as claims about distance from it—proximity in the courtroom era, when the Syrian's case rested on his closeness to his “neighbors on the European side of the Mediterranean”; separation in the directive era, when the case rested on the distance the existing box concealed. The White category is the axis of the system in both centuries: the largest box on the form, the one every other definition is written against, the one whose boundaries the courts, the commissions, and the budget office have each in turn been asked to police.
The census churn
The decennial form tells the same story in shorter chapters. The first census, in 1790, counted free white males by age, free white females, all other free persons, and slaves. The 1850 census added “mulatto” to the enumerator's marks, and mulatto stayed on the schedule, with interruptions, through 1920. The 1890 census carried the classification to its finest grain: “quadroon,” defined as one-fourth black blood, and “octoroon,” one-eighth or any trace, appeared for that census alone and were dropped in 1900. The 1930 census listed “Mexican” as a race for the first and only time, and its instructions to enumerators carried the one-drop rule in so many words: “a person of mixed white and Negro blood was to be returned as Negro, no matter how small the percentage of Negro blood.” The 1940 census removed Mexican and instructed its enumerators in the opposite direction: “Mexicans are to be regarded as white unless definitely of Indian or other nonwhite race.” A people entered the nation's racial taxonomy in 1930 and exited it in 1940, by instruction sheet. The Pew Research Center's history of the forms, updated November 3, 2025, summarizes the sequence in one line the desk adopts as accurate: the categories have changed from decade to decade, reflecting the politics and science of the times.
One drop, administered
Virginia built the era's most complete state machinery for the same work, and its file survives in detail. State law had long graded legal whiteness by fractions; a person with one-sixteenth or less of black ancestry could stand as white. The Act to Preserve Racial Integrity, passed in March 1924, replaced the fractions with an absolute: a white person was one with “no trace whatsoever of any blood other than Caucasian,” with a single exception—the so-called Pocahontas exception—for persons with one-sixteenth or less of American Indian blood and no other non-Caucasic blood. Birth and marriage certificates were to carry one of two races, white or colored. In 1930 the General Assembly completed the definition from the other side, defining a colored person as anyone with any ascertainable trace of negro blood: the one-drop rule as a statute. The act's administrator was Walter Ashby Plecker, a physician and eugenics advocate who served as registrar of the Bureau of Vital Statistics from 1912 to 1946 and had helped draft the law with the Anglo-Saxon Clubs of America. Plecker held that Virginia contained two races and no third thing; on that doctrine his office reclassified nearly all Virginia Indians as colored on their birth and marriage certificates, circulated lists of surnames, county by county, that it ruled belonged to mixed families, and instructed local registrars, hospital personnel, and school administrators to treat the named families accordingly. Records already issued were altered. The federal census counted 779 Indians in Virginia in 1930 and 198 in 1940. The marriage provisions of the act stood until Loving v. Virginia voided them in 1967; the altered paper stood longer—in 1987, at the request of Virginia Indians, the state vital records office agreed to restore “Indian” on the records of people Plecker's office had reclassified, and no Virginia tribe won federal recognition until 2018.
What the boxes hold
A minimum category is a roll-up, and the contents of the roll-ups are on the record. The 1997 White category spans the original peoples of three landmasses, and the Census Bureau's example entries for it place Irish, German, Italian, Lebanese, Near Easterner, Arab, and Polish in one line. The 2020 census collected more than 350 million detailed responses to its race and ethnicity questions, and its detailed file tabulates 300 detailed race and ethnic groups and 1,187 American Indian and Alaska Native tribes and villages—the plurality inside the boxes, published by the same bureau that administers the boxes. The Black or African American category gathers the descendants of American slavery, the peoples of the African continent, and the Caribbean into one reporting line; the new Middle Eastern or North African category is drawn geographically and takes in, in the Census Bureau's description, Arabic-speaking groups such as Egyptians and Jordanians, non-Arabic-speaking groups such as Iranians and Israelis, and ethnic and transnational groups from the region, such as Assyrians and Kurds—two dozen nations and peoples between Morocco and Iran, filed as one. None of this is concealed. The standards name their purpose in their own text: the categories exist so that statistics can be compared and civil-rights compliance reported against them. The flattening is the price of the comparability, stated in the documents themselves, and the peoples inside each box—their languages, religions, and histories—are the quantity the roll-up sets aside to get it.
Two other answers
The American system is one answer to an administrative question, and the century ran the alternatives at national scale. South Africa's Population Registration Act of 1950 ordered every person in the country classified by the state, and defined its top category by appearance and acceptance: a white person was one who “in appearance obviously is, or who is generally accepted as a white person.” Classification determined residence, schooling, employment, and political rights, and a machinery of officials and Race Classification Boards stood behind the register to correct it. By 1956, on the figures compiled by the South African Institute of Race Relations from the official record, the machinery had reclassified 18,469 people—1,182 reclassified as white, 9,642 as coloured, 7,645 as black—and the Government Gazette was still publishing annual lists of the reclassified in the 1980s, moving individuals among the categories in both directions by administrative act. The act was repealed in 1991. France runs the opposite answer. Its data-protection law of January 6, 1978 prohibits the processing of personal data that reveal, directly or indirectly, racial or ethnic origins, and when Parliament in 2007 voted to permit such processing for studies of diversity and discrimination, the Constitutional Council struck the provision: such processing, the Council held in decision 2007-557 DC, cannot “be based on ethnicity or race” without infringing the Constitution's guarantee of equality without distinction of origin, race, or religion. The consequence stands in France's own words to the United Nations Human Rights Committee in 2009: the Republic is “not in a position to provide statistics disaggregated by ethnic or racial origin.” One state classified everyone and reclassified them by the thousand; one state forbids the ledger; the United States revises its ledger by directive and is currently deciding, in the open, whether to keep it.
Every era's reason
The last entry in the record is the justifications, and they are uniform in one respect: each classifying regime described its system as an improvement, in the vocabulary its era supplied for virtue. The Immigration Commission's dictionary of 1911 presented its racial entries, temperaments included, as science placed at the service of legislation. The Virginia act of 1924 carried its rationale in its title—preservation—and its administrator was a public-health registrar applying eugenics, which the era's universities and medical journals taught as progressive science. The standards of 1977 and 1997 justified uniformity by its uses: comparable statistics and the enforcement of civil-rights law, a rationale printed in the standards themselves. The revision of 2024 justified its new boundaries by accuracy and usefulness, and by the data's role in measuring disparities and evaluating federal programs. The proposed rule of 2026 justifies removal by constitutional fidelity and by the integrity of the count. No document in this file describes its own classification as an exercise of power over the people classified, although classification is the entire content of each document; and no regime in the file describes the regime before it as having acted from the motives it claims for itself. The desk notes the pattern as a property of the documents, not of the men: the self-description is constant across systems whose classifications contradict one another outright, which makes the self-description the one line in each document that carries no information about the system it announces.
The record as it stands on November 14, 2026, is therefore this. The 2024 standards remain the governing federal standard for race and ethnicity data; their implementation has been postponed twice, their introduction on the government's largest survey has been withdrawn, and a proposed rule would bar the questions from the 2030 census, whose comment file closed November 2 with the Bureau's own impact analysis entered against the proposal. The 1997 map still governs the bulk of what the government actually collects, and under it Americans of Middle Eastern and North African origin are White, as they have been since 1977. Behind the current filings stands the longer file this report has printed: a statute of 1790 and the courtrooms that construed it, a dictionary of races, a registrar's surname lists, a one-drop statute, a classification board's annual gazette, a republic that keeps no ledger at all. In every period the classified are the constant and the classifications are the variable, and the hand on the pencil, in every period, belonged to an institution that believed its lines were drawn for the improvement of the society being lined.