May 22, 2020

An Enumeration of the Population: A history of the Census

By Margo Anderson

Editor’s Note: This article first appeared in the Winter 2010 issue of Insights on Law & Society as part of the last decennial census. The history is still a good one, and the author, Margo Anderson, is a renowned historian of the American Census, so we are thrilled to feature the piece for Census 2020.

We owe the American institution of the decennial census to the framers and the federal constitution of 1787. When the founders in 1787 decided to apportion seats in the new House of Representatives among the states “according to their respective numbers,” they invented a fundamental new instrument of republican government: the census. “We the people of the United States,” they decided, create the national government and are responsible for running and paying for it. To run the government, we need to choose our representatives.

There are elections, of course, so members of Congress elected from the states pass the laws and the President administers the government. But how is the President elected and how are representatives to be allocated among the states? The answer was the census, a periodic count of the population and consequent redistribution of House seats and economic resources to reflect the relative sizes of the populations of the states.

Even in the 18th century, the framers realized that counting the population would be difficult to do. It was put on a 10-year cycle since the framers knew that populations—and especially the American population—grew and shifted over time. Finally, the count needed to be done using uniform national procedures to ensure accuracy. As an essential element of the American political system, the census must be seen as fair and equitable to the variety of political, regional, and demographic communities of the nation.

The logic of the census system flowed from the experience and conceptions of the framers. They discussed apportionments based upon land assessments, other measures of wealth, and population. They agreed that political power should be allocated on the basis of population and that tax capacity derived from wealth. Population was much easier to measure than wealth, so population would be the apportionment measure. The new census and apportionment mechanisms of the federal constitution were thus a crucial piece of the Great Compromise among the large and small states, which made a national government possible.

But there was one fly in the ointment that would come back to haunt the framers in future years and that is still relevant today. That was the question of defining exactly who was part of the “population’’ deserving the right to political participation in the society and owing responsibility to pay taxes to the state. The discussion in the Constitutional Convention was to use the most expansive rule possible, including, for example, women, children, and the poor in the count, though they neither voted nor were necessarily responsible as individuals for taxes. The rub came when the framers considered how to treat slaves and Indians. Should the southern states, for example, be granted political representation for their slaves? At the time southerners considered slaves “property’’ for purposes of tax assessments but did not count slaves when they apportioned their state legislatures. Further, should Indians, who were generally considered outside the purview of the American polity and as members of foreign states, be counted for representation and taxation?

The census clause in Article I, Section 2, Paragraph 3 of the U.S. Constitution, continued: the “respective Numbers [of the population] ... shall be determined by adding to the whole Number of free Persons, including those bound to Service for a term of Years, and excluding Indians not taxed, three fifths of all other Persons.’’ The three-fifths compromise required the census to count slaves separately so each could be considered as three-fifths of a free person for apportionment. The second proviso eliminated “Indians not taxed’’ from the census altogether. Only people who came to be called “civilized Indians” were to be included in the decennial census count.

Demographic History

A census seems a fairly obvious tool to use to apportion political power among a set of constituencies. Yet the United States was the first nation in the world to institute a regular population count to apportion political power. The principles that political power was a function of population and that population could be measured were truly innovative in the 18th century and proved to be lasting as they were implemented in the 19th and 20th.

Moreover, the census might well have been a rather minor constitutional innovation were it not for the extraordinary demographic character of the American population. The United States has had one of the most heterogeneous and demographically dynamic populations in the history of the world. Over the past three centuries, the colonies and then the United States saw rapid population growth, major migrations, and
sharp demographic transitions—all in the context of a racially and ethnically heterogeneous population. The first
census counted 3.9 million people. The current land area of the country is four times the size of the nation in 1790; the population is almost 60 times larger. These dramatic trends over time make the decennial census and its apportionments major social, political, and intellectual events.

The First Census

The history of the census is also a revealing story about the history of the nation. James Madison proposed a rather elaborate census that included questions classifying the population by age, sex, and occupation, instead of the simple counts of the free and slave populations required by the Constitution. Congress rejected Madison’s proposal and settled on a more modest scheme. The first census law mandated the secretary of state to charge the United States marshals to count the population in six simple categories. For each household, the enumerator listed the name of household head, the number of free white males under 16, the number of free white males 16 and over, the number of free white females, the number of free colored, and the number of slaves. Even this simple count took 18 months to complete as the difficulties of counting the population in rural America emerged. When the first census recorded 3.9 million people, President George Washington thought the count was too low. But the count was credible enough to be used for apportionment and so Congress took up the second phase of the decennial census process: reapportionment.

The Constitution was silent on the method of apportionment. Soon Treasury Secretary Alexander Hamilton and Secretary of State Thomas Jefferson, the two leaders of the emerging political factions in the new government, came to be identified with two different methods of apportionment. Both were mathematically valid but had different effects on the distribution of seats in the House. A long debate emerged; Congress passed Jefferson’s bill. George Washington sought advice from Hamilton and vetoed the bill (using the presidential veto for the first time). Some weeks later, Congress passed a second bill— favoring Hamilton’s method, which Washington signed.

The lessons the framers took from their initial experience with census and apportionment were several. First, the census process “worked.” Further, the framers recognized that there were questions of “fairness” inherent in the apportionment mechanisms, which the Constitution did not address. Third, it was perhaps best not to tinker with the process. The “traditional” methods of the 1790s were thus reused for the next 50 years with minor modifications. Each decade, for example, Congress refined the age and sex breakdowns in the census and added a few new questions. By 1840, the six simple questions of 1790 had expanded to 70 or more.

Reforming the Census

In the late 1840s, Congress undertook to reform the census procedures and questions. In earlier decades, the census used the family as the unit of interest and reported few data on persons. For the 1850 count, however, Congress mandated an individual set of census questions for each person. It created a large, but temporary,
census office to tabulate the data. The 1850 census reform law had created six schedules for the statistics of population, mortality, agriculture, manufacturing, and social statistics.

By 1880, Congress recognized the weaknesses of the United States marshals and their assistants as the field staff for taking the census. The marshals were in charge of all field operations, including mapping their districts and appointing sufficient assistants to actually do the canvas. The accuracy of the enumeration relied very heavily on the abilities of the marshals—who otherwise worked for the Attorney General’s Office, not the Interior Department, which oversaw the census. The census superintendent had no real control over the marshals. There was no training, no means for teaching someone how to take the census except by the detailed instructions issued from Washington. Further, marshals were assigned according to the needs of the Attorney General, not the census, and everyone complained of uneven workloads.

Congress authorized the appointment of 150 census supervisors to manage field enumeration under the control of the census superintendent, compared to just 65 marshals in 1870. No enumeration district was to be larger than 4,000 people. The time for the enumeration was shortened from six months to two weeks in cities and one month in rural areas. Census Superintendent Francis Amasa Walker hired Henry Gannett from the Geological Survey to be census geographer. Gannett systematically mapped the country for the census, including many parts of the previously unmapped country west of the 100th meridian (present-day Dakotas, Nebraska, Kansas, and Texas).

Other major changes were prompted by the sheer volume of data pouring in on the census schedules. In 1880, census clerks still tallied the results for the census of 50 million people by hand. In 1890, census officials introduced machine tabulation of the census results. In 1902, the Census Bureau became a permanent federal agency and was housed in the Commerce and Labor Department in 1903. It remained in the Commerce Department when the two split in 1912.

The 1920 Apportionment Crisis

The chief controversy surrounding the census during the 1920s concerned the failure of Congress to reapportion itself after the 1920 census. For the only time in the nation’s history, Congress could not muster a majority to pass a reapportionment bill. The roots of the problem were several. First, Congress had decided in 1910 to stop the growth in the size of the House at 435 members. Thus, the 1920 reapportionment would be a true zero sum game; one state’s gain would be another state’s loss. Second, several mathematicians produced competing apportionment methodologies, which had the effect of producing different allocations for key states. Third, the demographic trends evident in the 1920 census results were not to the liking of the Republican majority in Congress.

The census results showed major shifts in population to the cities, to the far west, to places in the country populated by immigrants. The census, critics held, had to be wrong. So leaders proposed and failed to pass bill after bill until, by the late 1920s, key congressional leaders proposed an automatic mechanism to go into place after the 1930 census. This bill passed Congress with another provision that removed the mandate that required congressional districts to be compact and equally sized as nearly as possible within states. Congress, in other words, reapportioned power among the states but quietly acquiesced to rural malapportionment within states. States with growing urban populations would gain in the number of seats, but those seats would not necessarily be allocated to urban districts. It would be a practice that would lead to the reapportionment
revolution 40 years later and have a major impact on census taking.

Revitalizing the Census

The Great Depression dramatically changed the census and the American statistical system. The 1930 census was taken less than six months after the stock market crash. The Census Bureau found itself drawn into a contentious political debate about the scope and character of unemployment and thus the causes of the economy’s slide. The
bureau leadership was sympathetic with Republican President Herbert Hoover’s stance toward unemployment, and thus the bureau figures from the 1930 census and a special 1931 census of unemployment put the best face on what clearly was a dramatic rise in unemployment nationally.

In the spring of 1933, Franklin Roosevelt’s New Dealers came to town intent on remaking American government and ending the Depression. The Census Bureau came under particular scrutiny. Officials from Roosevelt’s alphabet agencies called upon the bureau to provide data on the socioeconomic situation of the population.

Congress built the grant-in-aid system to allocate tax money from the federal to state and local governments, and they sought population data on poverty, income distributions, and migration. Yet the Census Bureau of the early 1930s had yet to devise a credible measure of unemployment, had no data on income, and did not measure poverty. The New Dealers persevered and, by the end of the decade, had proposed the introduction of sample surveys to measure unemployment, reorganized the agency’s bureaucratic structure, increased the statistical training of employees, built a research unit, and analyzed and revamped classification systems. Many
of these innovations became part of the 1940 census. It included a sample long form for the first time, a housing census, and evaluation studies to systematically measure the level of accuracy of the enumeration, the tabulation and coding procedures, coding bias, and sampling error.

Discovery and Politicization of the Undercount

The issue of the undercount began to be framed precisely after a somewhat serendipitous natural experiment in 1940. The selective service registration of 1940 allowed demographers to compare the 1940 census counts of men of draft age (21 to 35) with the counts of men who registered for the draft. The census underreported about 3 percent of the men in the age cohort; the draft registration recorded some 453,000 more men than the census did. More significantly, though, was the finding that the level of the undercount varied by region and race. Some 13 percent of the black men of draft age were missed in the census. And black men registered for the draft in dramatically higher numbers in urban states than would have been expected from the census counts. By 1950, the Census Bureau knew it undercounted the total population by about 2.5 percent, and “nonwhites’’ by 11 percent.

Three separate trends merged to propel the census undercount onto the political stage in the mid-1960s. First, Congress continued to build the grants-in-aid system as a mechanism for allocating revenue to state and local
governments. Second, in 1962, the Supreme Court ruled in Baker v. Carr that malapportioned state legislatures were unconstitutional and opened the way for the decade of lawsuits that led to the “reapportionment revolution’’ of the 1960s. A series of cases followed that overthrew apportionments in other legislatures and in Congress. By 1964, the phrase “one man one vote’’ had entered the nation’s political vocabulary to define the new principle of legislative apportionment. Suddenly accurate census data for small geographic areas came to be of added importance, and the bureau had a strict constitutional duty under the equal protection clause of the Fourteenth Amendment to count everyone. Third, the Civil Rights Movement of the 1950s and 1960s used census data to frame many of its arguments about discrimination in the labor force, community participation, and access to housing in terms of underrepresentation of minorities in these areas of American life. If, for example, 30 percent of a local labor market were black, then, activists argued, true nondiscrimination would require blacks to hold roughly 30 percent of the jobs. Congress passed civil rights legislation that also relied on census data for administration and implementation. Under the Voting Rights Act of 1965, Congress created clear numerical tests of compliance: if a state used a literacy test for voter registration, and if voter registration or turnout was less than 50 percent of the voting age population of a jurisdiction, then the law presumed a violation of the Fifteenth Amendment. In such a case, the literacy tests were suspended, and the Justice Department could send federal registrars and election observers to monitor elections. The undercount ceased to be a technical problem of census field procedures; it became an explosive political issue.

New and Continuing Challenges

Since 1970, technological innovation and census accuracy have been major issues facing the Census Bureau. During this time, the bureau introduced technical innovations such as the mail census, a computerized mapping system and master address file of the nation, and electronic data distribution. In the past two decades the bureau has launched the ongoing national sample survey of the population, the American Community Survey, which replaces the census “long form.”

For much of the period, the bureau faced lawsuits from states and local governments that felt their populations were undercounted. By the 1990s the bureau developed adjustment methods that it planned to use in 2000 to improve the accuracy of the count. But the adjustment methodology encountered political opposition from Republicans in Congress who charged that counting methods that did not involve the “actual enumeration” of an individual were unconstitutional, and that the methods proposed to adjust or fix census errors would invite partisan manipulation of the data. The Supreme Court decided in Department of Commerce v. United States House in 1999 that the Census Act banned adjusted data for the congressional apportionment counts but left open the possibility of using adjusted data for redistricting and federal funds allocation.

This decade, as in the past, there will be controversies about the count. There are continuing issues of what are called “residence rules.” College students are supposed to be counted at their college location, but parents also tend to report them on census forms. College students are thus often double counted, and the question of which is the “right” location is not easy to resolve. A similar issue emerges for prisoners incarcerated in local jurisdictions, but whose home before incarceration was an urban district. There are implications for redistricting and funds allocations on such locational decisions.

The census will also face questions of whether some questions are an invasion of privacy, though in this country, such concerns have not fundamentally disrupted the count. This decade, the short form questions about name, address, age, sex, relationship to householder, race and ethnic status, and home tenancy are less likely to generate such concerns, though anyone with a potentially unlawful living situation, from an undocumented immigrant to a family violating zoning laws, could be concerned about filling out the form. The Census Bureau addresses concerns of privacy by assuring respondents their responses are confidential and are not accessible to any other government agency until public release 72 years after the census is taken.

Look for increased public discussion of the American “politics of population” in the months ahead. Watch to see how Americans respond, for the 24th time, to the census call to “stand up and be counted.” Finally, watch for the census results in late 2020 and early 2021 and our continuing national debate about who we are, where we’ve been, and where we’re going.

Margo Anderson is an American social historian and historian of statistics known for her studies of the United States Census and on the history of Pittsburgh and Milwaukee.