Just two days before the end of Trump’s reign, his appointed Director of the Census Bureau resigned following Bureau professionals’ resistance to his efforts to issue premature numbers in the waning hours of the administration. This was just the latest battle in the political warfare that enveloped the 2020 Census.
It’s not as if previous censuses avoided politics–they didn’t, as I discuss below–but 2020 was notable. For one, the Trump administration tried to add a citizenship question for the first time in 70 years, everyone understanding that its purpose was to scare immigrants, both documented and un-, into evading the count. The Supreme Court blocked that tactic. The administration also shortened the time available to complete the census even as inadequate funding and the Covid-19 pandemic made the work much more difficult. These moves would all produce underestimates of the population, especially in heavily Democratic districts and states. For the same purpose, the Trump administration asserted that House seats should be apportioned, for the first time ever, based only on the number of citizens and legal immigrants rather than of the number of “persons” as stipulated in the constitution (Art. I, Sec. 2).
But census politics goes back a long time–indeed, to the Constitutional Convention, where one of the North-South compromises ended up counting slaves as three-fifths of a person in the census, although, of course, without allowing slaves, nor women, nor Indians, nor the poor, even three-fifths of a vote. In late 1890, to take another example, the superintendent of the Census was compelled to write a ferocious defense against attacks on the validity of census, fending off charges about undercounting in New York (with all its immigrants) and overcounting in the South.
A review of recurrent political issues in the census puts this year’s chaos in perspective.
The Questions Matter
Race and ethnicity have been at the heart of many census controversies, which should be no surprise. Racial categories have appeared and disappeared over the centuries (mulatto and octoroon, for instance). Deciding whether and how to sort ethnic and racial groups is, of course, far more than a technical matter. On the ground, the counts determine political representation and the distribution of resources. How many Covid-19 vaccines end up going to Harlem depends in part on how many people you believe live in Harlem. And the numbers of, say, “Mexicans” (the 1930 census) or “Hispanics” (1970-on), affect political power.
In the 1920s, a Republican-dominated Congress refused to reapportion seats based on the 1920 census because the count showed too many immigrants and too many city people. The controversy led some who resisted the new reality to propose a constitutional amendment that would stipulate counting only citizens. Congress finally set up an automatic reapportionment system that started after the 1930 census, though it permitted very unequal districts within states that favored rural areas until the Supreme Court declared “one man, one vote” in the 1960s.
At a yet deeper level, census-formulated ethnic distinctions not only reflect the categories in a population, they also define them. Political scientists Evan Lieberman and Prerna Singh provide persuasive international evidence that where ethnic definitions and lines get drawn by a census, conflict along those specific lines is likelier to follow. In that vein, I recall being at an international conference to design a survey for replication in many nations. Some participants balked at using typical American questions about interviewees’ racial, ethnic, and religious identities, explaining that just raising such sensitive topics could put an interviewer in physical danger or at risk of arrest. Even in France, the census does not inquire about residents’ ethnicity or race and so “racial and ethnic statistical indicators do not exist,” making it hard, of course, to study racial inequities. Reify race and ethnicity or ignore race and ethnicity? Damned if you do and damned if you don’t.
The power of the census definitions explains why interest groups campaign to have themselves counted as distinct and numerous, as in the cases of American Indians and Latinos. For decades the census categories for native peoples varied–for example, sometimes counting Aleuts with and sometimes separately from lower-48 native peoples–and census-takers were told be strict about who qualified as Indian. In 1960, when self-administered questionnaires allowed people to classify themselves more easily, the number of Americans claiming indigenous heritage jumped by 50 percent. In an earlier post, I presented Cristina Mora’s account of how a lobbying campaign led to the addition of the “Hispanic” category to the U.S. Census in 1970, counting together Americans of Cuban, Mexican, Chilean, Puerto Rican, and other origins, thereby generating a large, single number. Even though more people prefer to label themselves by their ancestral countries than Hispanic or Latino (much less “Latinx”), “Hispanic” now carries political, commercial, and cultural clout.
A lobbying campaign also altered the Census’s descriptions of family structure. For most of U.S. history, the census-taker or questionnaire first asked who was the “head” of the household and then asked for the other household members and their relationships to the head. Married women could not be a head. (The Census corrected forms that listed wives as heads when a husband was present.) Feminist social scientists in the 1970s lobbied the Bureau against the assumption that there had to be a hierarchy with a head and the implicit corollary that the head had to be a man. The census changed to an egalitarian form in 1980 and in 2020 it asked, “If there is someone living here who pays the rent or owns this residence, start by listing him or her as Person 1.” The form then asks how everyone else is related to “Person 1.” The patriarchy–at least in the census–is gone. (And perhaps some day, “him or her” will become “them.”)
Controversy over undercounting was not just a last-century issue, of course. Censuses tend to miss about one in 10 residents, typically the poor, the minorities, the immigrant. Experts understand that statistically adjusting the data to correct for these missed persons would probably make the results more accurate (although that is debated), but that has never been allowed. Following the 1990 census, the G.H.W. Bush administration overruled the experts’ recommendations to correct the numbers; the adjustments would have increased the number of Democrat-inclined groups.
A new study of census history reveals that even the basic mechanics of counting can get caught up with politics and have political implications.
The Machines Matter
In a recent paper, “Census Technology, Politics, and Institutional Change, 1790–2020,” historians Steven Ruggles and Diana Magnuson describe how the recording and tabulating techniques of the U.S. Census developed over the centuries and the politics around that development.
The earliest censuses depended on local marshals hand-writing names onto lists and, in later years, onto prepared paper forms, tabulating them, and then sending the numbers off to Washington, complete with very many errors. Starting in 1850, local officials instead sent the filled-in forms to Washington to be counted by Census Bureau clerks–a massive operation made even more so by the continual addition of more questions. Introduction of the first punch cards and card-tabulation machines in 1890 saved much labor and also encouraged asking yet additional questions. In the early part of the 20th century, the Census Bureau itself made technological breakthroughs, including fill-in-the-bubble optical scanning, which eventually led to commercial spin-offs, notably IBM and Unisys.
Pressure to downsize the federal government starting in the 1970s led administrations of both parties to cut back Census personnel, forcing the Bureau to outsource the technical work. Major problems ensued. Vendors such as Lockheed Martin lacked the experience for the job, oversold their capabilities, charged a lot, and had buggy equipment and software, all of which slowed down censuses from 1990 on. Writing in mid-2020, Ruggles and Magnuson described yet new software development, further escalating costs, and a Bureau in turmoil. And then hit Covid-19. The privatization that started in the 1990s, they write, “was never about saving money…. The real agenda of privatization was ideological: the goal to prove that a Democratic administration” could also shrink government.
This, too, has consequences for how Americas are counted and described. The fewer the resources, the more complex the counting process, the tighter the budget, and the more rushed the canvass, the less the capacity there is to count the hard-to-count: the poor, the immigrant, the undocumented, and the marginal. And, in turn, the less help and less representation there is for the communities that harbor such people.
Sabotaging the census, unconsciously as the Clinton administration may have done in its pursuit of government efficiency or consciously as the Trump administration did, has effects not unlike the Congress’s decision in the 1920s to ignore the census results because they were too immigrant-filled.
MADE IN AMERICA 