Skip to content

Empire of the Census

America’s long history of manipulating its headcount for political gain

For all of American history prior to Reconstruction, slaves vastly outnumbered free citizens in Southern states. Fears that fueled Southerners’ refusal to give up their “peculiar institution” went beyond the economic and political; the white South was staring down a complete inversion of its existing social order if blacks achieved the full rights of constitutional citizenship. Since slave populations increased more rapidly than non-slave populations in Southern states—it was, after all, in the interest of the slave owner to encourage fecundity among his unwilling property—whites envisioned not only economic ruin but also the potential for racial Armageddon.

One group of voices, variously pro- or anti-slavery, counseled anxious proto- Confederate whites to remain calm: statisticians. Armed with census data, they argued that the solution was . . . emancipation? Ready for some seriously racist Nineteenth Century Scientific Rationalizing?

Horace Bushnell, a Northern theologian and statistical dabbler, was the first proponent of the argument that freeing the slaves was the only way for the white South to prevent itself from being dangerously outnumbered by people of African descent. Free black populations in Northern states, census data showed, did not experience explosive growth. White populations grew steadily. Slave populations grew rapidly. These differential growth rates obviously were cause for alarm in the South.

Bushnell concluded that white Southerners faced a choice: either find themselves numerically overwhelmed or accept emancipation on the theory that freedom would lead to the gradual disappearance of the black population. Of slaves, he argued:

At present they are kept from a decline in population only by the interest their masters have in them. Their law of population, now, is the same as that of neat cattle, and as the herd will dwindle when the herdsman withdraws his care, so will they.

African Americans, he concluded, naturally succumbed to laziness, vice, crime, indolence, and poverty when given freedom. Jesse Chickering, a fellow Northerner and statistician, concurred. Northern states’ census data showed “[the] tendency in [free black people], living in the midst of the whites and of the slaves, to degenerate, to dwindle away, and to become extinct as a race.” The numbers, in short, demonstrated that the black population would disappear without the nurturing hand of the slave master to sustain it. Southern legislatures, for their part, remained unconvinced. Alas, the bizarre argument was based on flawed data anyway.

The size of the slave population was of tremendous interest to white America, as it figured prominently in one of the nation’s foundational political compromises. The constitutional stipulation that representation in Congress shall be based on “the whole Number of free Persons, including those bound to Service for a Term of Years, and excluding Indians not taxed,[1] three fifths of all other Persons” incentivized Southerners to account for every possible slave and Northerners to keep a watchful eye on any Southern attempt to inflate the numbers.

Free blacks, on the other hand, were substantially undercounted in practice. The reason was obvious, as Frederick Douglass noted in 1851. “Colored people are, naturally enough, very suspicious of all white enquires into their numbers, and condition” so “when a white man enters the house of a man of color, opens his book, takes out his pen, and proceeds in the usual authoritative tone to make enquires, the feeling generally is, ‘that said white man is after no good’; and he is not very likely to receive very truthful answer[s].” Free blacks, existing in a purgatory between enslavement and full social equality, found that it paid to keep their heads down. They had nothing to gain, and potentially much to lose, from making a full accounting of themselves, their families, and their property to the government. Theirs was a precarious freedom.

Return of the Repressed

We can see from this brief history that there is a precedent for the consequences, in the xenophobic mindset of the modern American right, of adding a question about citizenship status on the 2020 census. When commerce secretary and cartoon plutocrat Wilbur Ross declared on March 26, 2018, against Census Bureau recommendations, that “Is this person a citizen of the United States?” should be added to census forms, it was not in a spirit of inquiry. The goal and inevitable outcome of the change are plainly obvious not only from logic but also from broader historical experience. Take a marginalized, fragile population, make explicit that The Government is interested in making lists of people like them (while regular Immigration and Customs Enforcement raid stories fill the news) and watch them disappear when the census-taker comes around. The odds that any noncitizen, even one with every conceivable legal blessing on their residency, will complete and return a mailed census form are below the odds that the 2020 Democratic Party platform will embrace full communism.[2]

One group of voices, variously pro- or anti-slavery, counseled anxious proto-Confederate whites to remain calm: statisticians.

This attempt to politicize the census has not gone unnoticed. Legal challenges have been filed, and major media outlets have dutifully given it coverage, often with an eye on the brazen manipulations of the present administration. Nevertheless, the history of the census, with the possible exception of a brief New Deal interregnum, has always been one of outright manipulation, politicization, marginalization, partisan gamesmanship, and hostility to minority groups. The same census once used to help the United States government round up Japanese Americans into internment camps cannot claim to be inherently benign.

Anything the Trump administration does to pervert the census would represent merely a return to the status quo for the majority of American history. Like so many other things—the welfare state, public education, progressive taxation—the idea that the census should be accurate and apolitical existed, if at all, only briefly in the short interval between the New Deal and the 1980 census. Everything since has been a mounting effort on the part of our economic and political betters to return the United States to the nineteenth century. Getting the numbers wrong is irrelevant so long as the data serve their partisan political purpose. What sense would it make to return to the economy of the Gilded Age without bringing back its politics as well?

Guns, Butter, and More Guns

The word statistics derives from the Latin statisticium, literally “of the state.” It reflects not only the central role basic data plays in the administration of government but also the immense difficulty of getting that data. Until the Industrial Revolution somewhat eased the burden, counting an entire population was a task beyond all but the most powerful monarchies.

Population counting is the most basic and important kind of state data collection, with examples dating to antiquity. Well-known historical censuses include those referenced in scriptures (e.g., the Book of Numbers) or the Domesday Book mandated by William the Conqueror in 1086. But evidence exists to suggest that Egyptian pharaohs were counting heads as early as 1500 BCE. Surely, William the Conqueror or the Han emperors were not interested in data for data’s sake. They wanted something every central government needs, namely reasonably accurate tallies of their two most important resources: wealth (including land and food production) and males of military age.

Censuses began as, and fundamentally remain, a way for states to determine how much war they can afford. To this end, the state needed an approximate answer to three questions: How much gold do we have? How many bodies can we press-gang into fighting? And for how long can we feed them? They undertook a cumbersome, costly census because the costs of guessing wrong about one’s own strength were high. Later, republican forms of government added another essential need for and function of census data: the allocation of political representation. In this area, the United States and its Constitution were trendsetters.

The Constitution was finished and ratified thanks largely to two major structural compromises; both were population-related. The first, the so-called Great Compromise, created a House of Representatives that rewarded more populous states with more legislative representation. The second, the Very Shitty Compromise, chose to punt on what to do with slave populations in this process by agreeing to count three-fifths of “all other Persons.”

The now-extinct process of direct taxation was also included in the Constitution not only as a means of generating government revenue but as a disincentive to game the system of population-based representation. Any state that fudged its population upward to increase its influence in Congress would be assessed higher taxes owed to the federal government based on that same figure.[3] Politically, this made the Constitution sufficiently palatable to thirteen states of varying size and both slavery factions. But it created the enormous practical question of how a nascent country, spread out across a continent and with a central government that existed only on paper, could collect comprehensive data on its own population. It was expected to take so long to do so that the Constitution itself apportioned the first Congress (based on earlier, dubious state-conducted censuses) rather than wait for the first census. This was in no small part because in 1790, railroads and telegraphs were still decades in the future. Travel and communication both occurred at foot speed—horse or human—and the lightly populated United States had few urban centers worthy of the name “city.” Enumeration meant literally walking the entire nation and making a headcount with pencil and paper.

As the federal government barely existed, new Secretary of State Thomas Jefferson assigned the task of conducting the census to the United States Marshals, a loosely knit confederation of about 650 quasi-law-enforcement officers spread widely across the landscape. The data they collected could hardly be considered reliable, accurate, or complete. But, for going door to door on foot across the whole country, with little guidance beyond “Try to count everyone, I guess,” it is a real accomplishment that the first census was completed at all. Things would improve. Eventually.

Book of Blunders

Nobody should take nineteenth-century data too seriously, including election results. Modern Americans ascribe a granite-like certainty and reliability to government statistics, and indeed much of what today’s bureaucracy churns out is among the most accurate data available. But that is a recent development. Early door-to-door census- takers, for example, took a one-shot approach to their work. If no one was home when a given door was rapped, that was it. Either the household size was estimated unscientifically (i.e. made up) or it was not counted. No follow-ups, no second canvass, no mail-in survey, no online supplement. Further, data was routinely lost, poorly sourced (substituting, for example, a town official’s data for a door-to-door count), or manipulated. Consider that the first census for which workers were given maps of the areas they were assigned to cover was 1850. And according to the Census Bureau’s own history, the first census that didn’t experience a major loss of data was 1840. “Major” is doing a lot of work in that sentence; it was not uncommon for whole states to return no data or fill coverage gaps with, at best, rough estimates.

The state needed an approximate answer to three questions: How much gold do we have? How many bodies can we press-gang into fighting? And for how long can we feed them?

The reasons for this were not merely practical—the difficulty in executing a nationwide count in a preindustrial society—or bureaucratic. The early censuses were conducted in a political environment still deeply paranoid about the reach of federal power. Just as there was little standing Army—the military was built hastily for conflicts and then disassembled at their conclusion—the census office was not permanent. It was disbanded after each census, then recreated shortly before the next one.

Nor was the task professionalized or allowed to benefit from the accrual of bureaucratic experience and knowledge until the formation of the permanent U.S. Census Bureau in 1902. In fact, this absence of a permanent bureaucracy served a crucial function for the Congress of the era: it left a hole for the census to fill with short-term patronage. Congress, in other words, wanted a precision census much less than they wanted a decennial chance to distribute cushy jobs to every hanger-on in every party machine across the country. As the United States exploded in size throughout the nineteenth century, the number of people required to prepare for, execute, and process the census rapidly grew into the thousands. And prior to the Progressive Era civil service reforms, all such opportunities were seen by an eager Congress as a way to solidify support by handing out the goods.

That said, there was no real need for precision in early censuses. As long as they were broadly accurate—that is, they more or less reflected the relative size of each state—they were sufficient to serve their primary constitutional purpose of apportioning House seats. If there was nothing to suggest that the conduct of a given census wildly inflated one state’s population or short-changed another’s, nobody cared about the finer details. Knowing precisely how many people lived in Warwick, Rhode Island, interested no one.

Until the size of the House was fixed at 435 in 1929, apportionment was a bizarre, subjective, and highly politicized process. The size of the House was set—by the House itself—after each new census was certified, and the guiding principle used was that no state should lose seats. Each decade a so-called Apportionment Act was passed after the House devised a way to accommodate growth in some states relative to others (and the addition of new states) without causing any of the established states to give up a seat. So the House just kept getting bigger. If a similar system were still used, the census would not ruthlessly take a seat from Michigan every ten years and give it to Texas; the House would simply increase in size to whatever figure allowed Michigan to tread water.

Under those circumstances, the census could be accurate like the judging of Olympic gymnastics is accurate; as long as it ranked the contestants in the desired order in the end, the precision of the numbers (was that routine a 9.4 or a 9.3?) did not matter overmuch.

Revenge of the Nerds

When FDR took office in 1933 with an army of ardent New Deal enthusiasts and a lot of big new ideas about what government could and should do, they all quickly realized that an accurate census was essential and the country didn’t have one. They wanted to know, for example, exactly where to target federal relief programs, and in what amounts. Social Security, farm relief, and unemployment support were all key programs that could not be carried out effectively without having the highest possible level of geographic precision in population data.

Large scale economic intervention required previously unmatched state capacity and bureaucratic skill, and the need for good data quickly became fundamental. Belief in a professionalized, permanent bureaucracy relatively free of politicization was already established in much of Europe and, while a key component of earlier Progressive Era reforms, did not come to full fruition until the New Deal expansion of federal responsibilities.[4]

Alongside the New Deal economic and political shifts necessitating more precise data, a paradigm shift was under way. The idea that data should be as accurate as possible just because fit the zeitgeist of the moment. Suspicion of federal power had long been a feature of American politics. The modern American ironclad certainty that anything any government has done or is doing is incompetent and unnecessary, on the other hand, had not yet become an article of faith.

Post–New Deal Democrats, buoyed by the immense organizational and bureaucratic task of conducting the Second World War, carried forward the Good Government ethos of the man who had returned them from the post–Civil War political wilderness. It would be naive and incomplete to suggest that political considerations disappeared from governance, but there was a real revolution from the pre-FDR practice of any bureaucratic post being awarded to politicians’ literal drinking buddies or party hangers-on irrespective of basic skills necessary to do the job. At times to a fault (for example, Kennedy’s Whiz Kids’ data-driven insistence that the Vietnam War was a good idea, and winnable) belief in Good Government was a core liberal principle.

Out of Our Census

Bureaucratic change is slow, much slower than political change. The 1980 census, not to mention the ascension of St. Ronnie in the same year, may have signified the end of generalized belief in the usefulness of a capable federal government with sufficient resources doing essential tasks, but the bureaucratic atrophy engendered by the renewed “government is the problem” ethos has unfolded slowly. Not only is government losing capacity it once had, but pre-twentieth-century political beliefs are sufficiently popular now to leave questions like “Should the census be accurate?” once again open to debate.

The census under Trump is, like all other aspects of the federal bureaucracy, subject to the limitations of unfilled positions, insufficient resources, and political meddling of the sort one might expect if the officials in charge were plucked at random from a Facebook comment section. Commerce secretary Wilbur Ross can be distinguished from your least-favorite uncle only inasmuch as he is wealthy.

The census is defined clearly in the Constitution as a decennial count of all Persons—not citizens, as when the Constitution means citizens, it uses the word “Citizens”—in a given geography. State and local governments, in theory, want to increase their population count any way they can, as countless important sources of federal funding and aid are based on that figure.

If no one was home when a given door was rapped, that was it. Either the household size was estimated unscientifically (i.e. made up) or it was not counted.

But in this era, the plutocrats (and their lackeys) who want the figures to be wrong are holding power. To them, doing shoddy work that could readily be done well is a feature and not a bug. What’s more, they are happy to take the cheap, short-term gains of enthralling the worst people in America with performative immigrant bashing and racism, consequences be damned. And if the resulting census data overrepresents and inflates the relative importance of the easiest people to count (old white people who have not changed addresses since the Ford years) in the short run, all the better. Like the small, provincial, don’t-know-don’t-care men they are, the people swept into power on Trump’s tidal wave of raw sewage have no interest in anything beyond the crassest kind of self-interest. They reek of the CEO world from which they came, perfectly content to see the United States as another company to strip of its assets before they line their pockets with the proceeds and walk away.

There may be schadenfreude at the end of this saga, as the impending large undercount of people of color (particularly noncitizens) will affect some deeply red states in addition to hated liberal bastions like California, New York, and Illinois. Texas is likely to lose congressional representation if the census not only includes the citizenship question but is hamstrung by its top brass. Many people are extremely difficult to find and count even when the Census Bureau uses every tool at its disposal to do so. With anything less than an all-out effort using the best methods available, the results will continue to exaggerate the historical pattern of undercounting.

Those undercounts will affect governing long past Trump’s departure. Data from the 2020 census will form the foundation of government work across the board until an opportunity to fix it in 2030 arises. At this rate, though, it’s not a certainty that future efforts won’t suffer the same fate. This is a nation, after all, where the comfortable demand twenty kinds of instant replay to ensure that the outcomes of their football games are correct but couldn’t care less if the data underlying all government programs and services is accurate.

It’ll take a while for them to feel the consequences of this effort to score cheap political points by trying to scare Latinx people away from the census. As ever, the insufficiently affluent—those who rely on things like public schools, roads, government-subsidized health care, and more—will feel them immediately. And that pleases the right to no end.

The sad thing is not that the 2020 census is likely to be less accurate than it could be; no, it is that a stupendously wealthy, industrialized, modern nation does not want to know how many people live within its borders. Not wanting a census to be accurate is the nation-state equivalent of not wanting your bank statements to tell you how much money is really in your account. In time that may prove to be the tidiest metaphor for this entire era.


[1] This fragment represents the contemporary legal status of Native American tribes, who were treated as independent nations not subject to United States jurisdiction, a legal fiction that held no longer than it took for westward expansion to begin.

[2] A conservative estimate, based on other Census Bureau projects that include a citizenship question, is depressed response rates of 6 percent among households with one noncitizen member compared to all-citizen households. See Brown et al. 2019. “Estimating the Potential Effects of a Citizenship Question to the 2020 census.” IZA Discussion Paper 12087. Bonn, Germany: Institute of Labor Economics.

[3] Direct taxation, described in Article I, Section 9, Clause 4, was a cumbersome system by which Congress would fund the government through periodic “Revenue Acts” assessing states a flat rate of taxation based on population. This occurred irregularly for a century as, along with foreign tariffs, the primary means of funding the national government. Direct taxes of this type were prohibited by the Supreme Court in Pollock v Farmers Loan and Trust Co (1895) and replaced in 1913 with the passage of the 16th Amendment permitting income taxation.

[4] From 1933 to 1945 the federal workforce grew from around 500,000 to 3.5 million, its all-time peak. Today the civilian bureaucracy employs about 3 million workers, most of whom, at the time of this writing, were not getting paid because the president is an asshole.