The following excerpt is from the report “Politics and Science in Census Taking,” by Kenneth Prewitt, published by the Russell Sage Foundation and the Population Reference Bureau. This report is one of several in the new series The American People, which sets the results of Census 2000 in context and collectively provides a portrait of the American people in a new century. Each report is written by an author or team of authors selected for their expertise with the data and their broad understanding of the implications of demographic trends. Reynolds Farley and John Haaga are the series editors. For information on The American People, go to www.prb.org/AmericanPeople.
Government and Democracy, Politics and Science
(November 2003) Mention the word “census,” and what comes to mind is a dull counting project that the government carries out from time to time. Ask why a census is taken, and most Americans will vaguely reply that the government seems to need all these numbers. A few might add that it has something to do with who goes to Congress and even with how federal monies are spent.
Not many Americans know that the census is required by the Constitution and that since 1787 it has protected basic democratic principles. Many will be surprised to hear that there is an intense politics of census taking and that in recent times these politics have turned sharply partisan. Americans will be equally surprised to learn that the census is the nation’s longest continuous scientific project. Dull and technical though it may sound, the census is a drama at the very center of our political life.
This drama has four subplots: providing the government a read on society, insisting that democratic principles be honored, determining political winners and losers, and applying statistical science on a scale nowhere else matched. In the best of times, these four subplots flow comfortably together. Sometimes, though, they don’t, creating tension and confusion. Such was the case in the 2000 Census. Government goals were pulled in different directions. Principles of democratic fairness were in jeopardy. Issues of racial justice were contested. The politics were intense and partisan. Census science was vigorously debated, even attacked as unconstitutional. The Supreme Court had to weigh in, twice in two years. This report is about these conflicts, using them to show how policymaking, democratic principles, partisan politics, and science come together in the census.
We briefly introduce each of the subplots before subsequent sections reveal how, when, and why they link in a single narrative.
In the fourth book of the Old Testament, aptly titled Numbers, God tells Moses: “Take ye the sum of all the congregation of the children of Israel, after their families, with the number of their names, every male … from twenty years old and upward.” Moses is being told to take a census of those “able to go forth to war.” There is also a census in the New Testament, central to the nativity story that brought Mary and Joseph to Bethlehem, where they were enumerated in order to be taxed. These biblical stories describe ancient practices in which census taking was used by the state to protect the security of its territory and its citizens and to raise revenue. Military conscription or taxation, and often both, are linked with census taking from the beginning of recorded history.
Think of the census as a map — a map of the society rather than of the territory. Both maps and censuses tell us what is out there. An ambitious 17th century annual census is proposed to France’s Louis XIV in these terms: “Would it not be a great satisfaction to the king to know at a designated moment every year the number of his subjects, in total and by region, with all the resources, wealth and poverty of each place? … [Would it not be] a useful and necessary pleasure for him to be able, in his own office, to review in an hour’s time the present and past condition of a great realm of which he is the head, and be able himself to know with certitude in what consists his grandeur, his wealth, and his strengths?”1
The French king did not get this ambitious census, but modern governments do. In the United States, the 2000 Census form asked questions ranging from veteran’s status to place of birth, from age and race to education and ancestry, from distance traveled to work to the costs paid for home heating fuel.
These are not matters of idle curiosity. Every question asked in the U.S. census connects to a specific government program or purpose: locating medical services for veterans, enforcing the Voting Rights Act, monitoring the changing skill level of the work force, planning transportation networks, designing energy policies.
Imagine that you had the task of targeting educational support to children in poverty or medical services to physically disabled veterans. The size and geographic location of such specific subpopulations are provided by the census and its many derivative surveys. Every year, census numbers help determine where approximately $200 billion in federal funds are spent for welfare, education, health, transportation, and dozens of other programs.2 To pay for these programs, government turns to the census to assess the size, wealth, age structure, and employment patterns of the tax-paying population and designs a taxation system accordingly.
Think of the census as a huge report to the government, portraying what is out there so that laws and programs can be anchored in information that is systematic, comprehensive, current, and objective.
So far, so good, but there is more. Knowing what is out there can lead to a variety of policies, not all of them benign. Governments around the world, for example, generally want to control who enters or leaves their country. In addition to border controls, visas, and the like, they use immigration statistics to determine how many of what kinds of people from what regions of the world qualify to enter the country. A government may also want to influence the overall size of the population, and will use fertility rates and statistics on the age structure of the population to design family planning policies.
The boundary between routine population policy and abusive exercise of power by a government is easily breached, and census history offers instances of the latter. Fertility regulation and forced sterilization in China and India come to mind. Population relocation is another example, and the United States provides a dismal illustration with the internment of Japanese Americans at the beginning of World War II. In this instance, census data helped the government quickly round up large numbers of citizens of Japanese origin, especially on the West Coast, who were said to pose security risks and thus were isolated in guarded camps.
In its darkest misuse, census information has facilitated genocides. Small-area tabulation of a Dutch census of religions taken in 1930 was later used in dot maps of Amsterdam to indicate density of the Jewish population, making it easier for German occupation forces to deport them. The 1939 German census went further, being used to identify specific Jews and Gypsies targeted for concentration camps.
These perversions are few in comparison with the majority of cases where the public gains from fairness and policy efficiency because a census draws an accurate portrait of the population. But like any powerful instrument of government, a census can be used for ill as well as for good.
Insisting That Democracy Be Fair
America had its first national census in 1790, for a purpose that was new to census taking. America’s Constitution took on the historically unprecedented task of establishing a government based on the popular will. Perplexing questions faced the writers of the Constitution. What theory of political representation would work for the new republic? How could this republic avoid the seductive temptation of empire?
The Constitution establishes a government based on democratic representation — a few citizens are elected by popular vote to represent the will of the many. This elementary principle of democracy leads to a practical question: What groupings of the population will elect representatives? One possibility is an at-large election in which a long list of candidates is presented to voters across the country, and candidates with the most votes then represent the entire population. But the nation’s founders were seeking a way to bring elected representatives closer to the people being represented — that is, elections in which specified groupings of voters would each select their own representative. What principle might define these groupings? The Constitution writers could have divided the population occupationally — farmers, manufacturers, tradesmen, professionals — with each occupation electing its representatives. Or perhaps religious groupings — Anglicans, Congregationalists, Quakers, Methodists — could have served as the basis for representation. Such alternatives were not seriously considered.
Political representation in America is based on territory. Representatives are elected by voters from a politically defined area. This was the principle most familiar to the Constitution writers when they gathered in Philadelphia. In the decade between independence and the Constitutional Convention, each of the 13 original states had its own political identity, rooted in its economy, its religious tradition, its size, and how it had been governed during the colonial period. Following independence the country was governed under the Articles of Confederation, which gave one vote to each state in determining broad policy for the country while also accepting that most policy — including taxation, education, and criminal justice — would be made at the state level. This allowed the Southern states to continue their plantation economy based on slave labor while the Northern states could turn to small-scale farming and commerce. It allowed for differences in religion and cultural outlook; for instance, Puritanical Calvinists concentrated in Massachusetts, Quakers in Pennsylvania, Catholics in Maryland, and Anglicans in Virginia.
The delegates to the Constitutional Convention in 1787 were sent by their respective state legislatures, mostly composed of leaders fiercely protective of state interests. The delegates to Philadelphia knew that a stronger federal constitution to replace the weak Articles of Confederation would be politically acceptable only if state interests were clearly recognized. The result was political representation based on territory: Voters in each of the 13 founding states would elect their representatives to represent their interests in the new Congress.
Immediately there was an issue. How many representatives should be elected from each state? The smaller states wanted every state to have equal voting power. The more populous states cried “unfair,” arguing that power over national policy should be proportionate to the size of the state’s population. In perhaps the most important compromise of the Constitutional Convention, this sharp controversy was sidestepped by creating a two-house Congress. In the Senate, each state has equal voting power; in the House of Representatives, voting power is proportionate to the population of the state.
The first Congress would have a total of 105 House members, with large states getting more members than small states. This posed the question of how to be fair in determining the exact number of members each state would get. The first step was to take a census to establish the share of the total population that lived in each of the 13 states. The nation’s first census took place in 1790, and with the results in hand, an allocation formula written by Thomas Jefferson fixed the number of seats each state would have in the House of Representatives, ranging from one for Delaware, the least populous state, to 22 for Virginia, the most populous.
These differences in state-by-state power in the House put firmly in place the fundamental principle that representation is based on residency and is proportionate to population size. This principle has guided American democracy for more than 200 years. But if the 1790 Census solved the initial problem of democratic representation, why did the Constitution require that a census be taken every 10 years?
Admitting New States to the Union
The boundaries of the United States were not fixed by the War for Independence, which resolved only that 13 British colonies were free to govern themselves. Nor were the boundaries fixed by the Constitution, which resolved that the 13 original states would come together in a federal government. Not at all clear was the status of Americans who lived outside the boundaries of these 13 states.
This uncertainty had to be addressed. Already restless people were leaving the 13 states, crossing the Appalachian mountain range, spreading into the Ohio Valley and down the Mississippi to the Gulf of Mexico. Among the drafters of the Constitution were visionaries who imagined a nation that might one day reach across the sparsely settled continent even to the Pacific Ocean. Territorial expansion was much on the mind of the new nation.
And what was to be the political status of these Western territories — so rich in land, minerals, furs, and forests? They could be annexed as colonies, and their resources exploited for the benefit of the Eastern Seaboard states. Alternatively, the Western territories could be encouraged to form themselves as independent states that would join the Union on equal footing with the original 13. Principle won out over economic expediency, as a majority of the framers felt that an Eastern Seaboard empire with Western colonies violated the principles for which they had waged the War for Independence. Empire was rejected in favor of allowing the new territories, as quickly as they were settled, to apply for statehood with powers and responsibilities equal to the original states. The key phrase is “as quickly as they were settled”; a decennial census was needed to determine when a territory was sufficiently populated to become a state.
The difficult issues of designing a democracy were hardly all resolved at the Constitutional Convention. American political history is the ongoing story of democratic principles evolving in response to conditions unforeseeable in 1787. Were we to recount that entire history, we would see that the census was often central. Our subject, however, is primarily the recent period, and especially how democratic principles were invoked in arguments about the census in 2000.
Determining Winners and Losers
The size, geographic distribution, and characteristics of the population are linked to power, to money, to civil rights. With such high stakes, there will be sharply competing interests in census results.
In the late 20th century, a vigorous political debate emerged about how census taking should address an old problem, the difficulty of counting everyone in the census. The problem was not new: After the census of 1790, the first presidential veto in American history set aside the census apportionment formula prepared by Alexander Hamilton, which favored commercial and manufacturing interests, opening the way for Thomas Jefferson to propose a formula that benefited agricultural interests. Two hundred years later, census methods were being debated before the Supreme Court in a case brought by congressional Republicans against the Democratic president, Bill Clinton. The issue again was how competing interests could be advanced by population numbers.
A census is unfair when some areas or groups are more fully counted than others, because the less well counted do not get their rightful share of benefits — such as legislative seats or public monies. If racial minorities are among the less-well-counted groups, as they are in the United States, issues of social justice and equality are joined to the census. There can be no account of America’s census without looking at the politics of numbers — how many, how many where, how many of what groups.
We have not said much about census taking itself — how the census is conducted (see below, “Major Phases of the Census”) and, more important, how accurate it is. It is not easy to enumerate a population that is mobile, busy, and easily distracted; that is often indifferent to its civic duties; that is sometimes resentful toward government and not inclined to be cooperative; that speaks dozens of languages and is spread from the edge of the Bering Sea to the tip of Florida; and that is mostly ignorant of what is at stake in the census numbers.
A few dozen U.S. marshals carried out the first census. The most recent census involved nearly a million employees, making it the largest peacetime mobilization in American history. Census taking could not have kept pace with the enormous changes in the country and the population over the last two centuries without itself changing. As a vast scientific and technical enterprise, each census works to correct the flaws and build on the accomplishments of the one that preceded it, making it the nation’s longest continuous scientific project.
The Census Bureau has been home to many scientific and technical accomplishments. Modern computing traces its origins to the late 19th century, when a bureau employee, Herman Hollerith, invented a primitive punch-card device and then created the first automated data tabulation system. (He later helped found IBM.) The Census Bureau was the first government agency to use a mainframe computer. In the 1930s the bureau used modern sampling theory for large-scale surveys, paving the way for today’s vast polling industry. The science brought to bear in the 2000 Census involved mathematical statisticians, geographers, demographers, psychometricians, sampling theorists, information specialists, and computer scientists, all joining in the demanding task of accurately counting the population.
Accuracy is a technical and scientific matter, but in the case of the census it is also a term tangled up with the larger issues of governing and democracy. Power is allocated, funds are distributed, and civil rights guaranteed on the basis of census numbers. But what if the numbers are inaccurate? For better or for worse, census accuracy is where the politics and the science collide, and how the resulting conflicts are resolved carries into government policy and democratic practice.
Major Phases of the Census
The old saw that the devil is in the details is especially true of census taking. This note sets out the major phases of the U.S. census as practiced in 2000. Readers will find it helpful to refer to this outline as different phases of the census are discussed in subsequent sections.
Step 1: Construction of an Address File. The U.S. census is based on residency. It starts with a master address file that is assembled from post office delivery addresses; block-by-block canvassing by census workers; the updating of addresses by local governments, adding new construction; and the filtering out of duplicate addresses. The master address file, of approximately 120 million addresses, is the control against which all subsequent census steps unfold. If it is incomplete, the census is incomplete. If it has unrecognized duplicates, the census will have duplicates.
Step 2: Form Delivery. A census form is delivered either by mail (the vast majority) or by hand (where there is no residential mail delivery) to every address, with instructions to return the form by mail. This is the critical mail-out/mail-back phase of the census. Lack of cooperation — that is, low levels of mailing the census form back — drives the costs of the census up and the quality of census data down. There are two census forms. The seven questions on the short form are asked of every household. An additional 47 questions are asked of a sample of one in six households on what is known as the long form.
Step 3: Appeals for Public Cooperation. Auxiliary operations are closely associated with the mail-out/mail-back phase — an advertising campaign, help from organizations that partner with the Census Bureau, mobilization by local leaders, and heavy media coverage — all designed to increase mail response and cooperation with follow-up operations.
Step 4: Nonresponse Follow-up. Not every form is mailed back, requiring several hundred thousand census takers, often called enumerators, to follow up with nonresponding households. Approximately one-third of the addresses on the master address file had to be contacted in this labor-intensive way in 2000.
Step 5: Quality Control. An independent sample of the population was used to estimate the rate of undercounting or overcounting of specific subpopulations, such as renters, children, or racial minorities. This Accuracy and Coverage Evaluation (A.C.E.) of Census 2000 was by far the most ambitious part of the quality-control operation. Others focused on checking census coverage — who was missed, who was erroneously included — and on data quality.
Step 6: Data Capture. Census answers — mostly tick marks in survey-form boxes — are converted to electronic records for tabulation and reporting. In Census 2000, technically sophisticated optical scanning methods were successfully used in digital data capture.
Step 7: Dissemination of Results. The first instance of announcing and disseminating census results is the all-important announcement of state populations, which determine the allocation of seats in the House of Representatives. By law, this census result is delivered to the president, who then transmits it to the Congress within nine months of census day (April 1 since 1930). The next major census product, due three months later, is the redistricting data file detailing the age, gender, and ethnic-racial breakdown of the population to the block level. This census product is used to redraw the geographic boundaries of congressional districts, state legislative districts, and county and city election areas to reflect population changes since the previous census. Subsequent and much more detailed census products are released as soon as feasible and are posted on the Census Bureau website.
Step 8: Evaluation and Planning. Even as one census is underway, its procedures are being evaluated with an eye to designing the next census. Studies embedded in one census point to improvements for the next. Did the advertising campaign work as intended? Did asking local governments to update the address file solve more address problems than it created? Could absolute confidentiality be protected if people filled out the form online? Could greater use of administrative records, such as school enrollment, improve coverage? These and dozens of other issues were part of Census 2000 evaluation studies. Studying such questions during the census process is far preferable to conducting less realistic studies outside the census environment. In this regard, conducting a census is like sending a spaceship to Mars. All the tests, plans, and experiments cannot substitute for actually doing it.
Kenneth Prewitt is the Carnegie Professor of Public Affairs, School of International and Public Affairs, Columbia University. He was director of the U.S. Census Bureau from 1998 to 2001.
- James Scott, Seeing Like a State: How Certain Schemes to Improve the Human Condition Have Failed (New Haven: Yale University Press, 1998): 11.
- Committee on National Statistics of the National Research Council, Statistical Issues in Allocating Funds by Formula (Washington, DC: The National Academies Press, 2003).