Credit: Amber Huff

In 1980, toward the end of a press conference on the state of the census count, Dianne Feinstein, then the mayor of San Francisco, turned toward Census Bureau director Vincent Barabba with a harsh warning: if he didn’t recognize her requests, she said, “We may see you in court.”

That year, it was just about the most ordinary thing that could have been uttered in a public exchange between a mayor and the head of the bureau, who would be on the defending side of no fewer than 52 lawsuits by the end of the decade. Like a growing number of mayors across the country, Feinstein was frustrated by the persistence of an undercount and the appearance that the bureau’s efforts to decrease disparities hadn’t moved the needle much.

In 1973, the bureau had announced that 5.3 million people were not included in the final tally for the 1970 Census. The overall undercount for the nation that year was 2.5 percent—but the differential undercount, or the difference between the rate of undercount for Black people and white people, revealed serious racial bias. Black people were undercounted on average at the rate of 7.7 percent, four times the rate for white people. (Hispanic was not fully listed as an option for ethnicity until 1980, though Latinos were also likely to be missed at a disproportionately high rate, the bureau later conceded.)

There had never been a year without a significant undercount since the first census, in 1790, when the three-fifths clause constitutionally mandated the undercount of Black people by 40 percent; the bureau had not counted Indigenous people at all until 1840. But hundreds of years after the repeal of the clause, and 17 after the height of the Civil Rights Movement, it was shocking that such clear racial bias remained in the country’s most important data set. And with recent policy changes increasing the importance of census data to the apportionment of congressional seats and the distribution of federal funds to cities and states, the differential undercount started to be viewed as a serious threat to the voting rights of minority groups.

In the years leading up to the 1980 Census, the bureau conducted an apology tour of sorts. “We didn’t do as good a job counting black people as we should have,” Barabba conceded to the New York Times in 1974. Under his tenure, the bureau added programs and increased investments at federal and regional levels to increase census outreach in cities, where it was most likely to miss people, and made inroads with Black and Latino leaders, including Bobby Seale of the Black Panther Party and other important radicals, in the process of creating new Black and ethnic advisory committees.

While many applauded the bureau’s gestures toward inclusivity, a growing group was critical of the bureau’s plan to eliminate the undercount by putting more money and resources into the same methods that were missing millions. Instead, they posed a question that would open up a 20-year political, statistical, and legal debate about equity in the census: If Census Bureau statisticians were capable of calculating the number—and the demographic information—of the people they had failed to count, couldn’t they add those people back into the data for the final tally? And if they could, shouldn’t they?

In November 1979, as preparations were underway to conduct the most expensive census to date, the newly formed Illinois Council for Black Studies organized a conference at the University of Chicago to discuss the history and political impacts of the census undercount on Black and other minority communities. It was the cusp of winter and snow was starting to fall, but still around a hundred people had made the trek, some from down the road in Woodlawn, some from across the city, and others yet from as far as Mississippi and Washington, D.C.

To Abdul Alkalimat (who is also known by the name Gerald McWorter), a founding member of ICBS and a professor emeritus of African American Studies at the University of Illinois at Urbana-Champaign, it seemed appropriate that a Black Studies council, which followed the motto of “Academic Excellence and Social Responsibility,” should host a reform-oriented forum not just between the bureau and scholars but also organizers and the general public.

“There was a major concern that the undercount of Black people would diminish the possibility of Black self-determination,” he said when we talked via Zoom in late July. Wearing a beige T-shirt and wire-rimmed glasses, Alkalimat swiveled slightly in his office chair. “[The census] became an issue that we were trying to link to Black Studies, which was becoming the most successful project of the Black Power Movement.”

In the tradition of the field of Black Studies, the conference, “Black People and the 1980 Census,” crossed disciplinary and political lines. A wide range of participants had been invited to share their perspective on the census and how its skewed picture of the country’s population affected their work and communities. That list included organizers representing the South Shore Housing Center, the Woodlawn Organization (TWO), MALDEF (the Mexican American Legal Defense & Educational Fund), and Chicago’s Native community, to name a few; Chicago Public Schools deputy superintendent and chief operating officer Manford Byrd; Illinois First District representative Bennett M. Stewart and Mississippi senator Henry Kirksey; and Census Bureau deputy director George Hall (who attended in place of Barabba, who had fallen ill), among others—in short, a “who’s who” in Black politics, scholarship, and Chicago organizing. For a census gathering, it was monumental.

It was clear that little was left untouched by incomplete census data: school district funding, hospital funding and distribution across the south and west sides of the city, affirmative action quotas, employment and unemployment statistics, the fight against gerrymandering in Mississippi, research on the economic and social barriers that Black women face.

At the center of discussion, however, was a bigger question. Census data had always been vital for House apportionments, redistricting, and funding. But over the last decade, the amount of federal funding distributed to states and cities via population data had increased. And the One Person One Vote Rule of 1964 now required that congressional districts be drawn to contain roughly equal population size to ensure that all residents in any given state had equal voting power. The rule was intended as a protection against the disenfranchisement of Black and minority communities, but, in light of growing public consciousness of the severity of the undercount, some questioned whether that promise of equal representation could be upheld.

This was a serious concern in major cities, where more than 75 percent of the country’s Black population lived by 1980. In a presentation on the census undercount of Black people as the “new disenfranchisement,” ICBS executive director and Northwestern political science professor Ronald Bailey wrote that it was possible that in the 1981 redistricting process, one of the three majority Black districts in Chicago could be expanded or combined with another district. He was particularly worried about the First District along the south lakefront, which had seen a dramatic ten-year population decrease of 70,000, almost a quarter of the city’s total population loss that decade. How much of this declining population loss was compounded or overstated by the undercount?

Bailey calculated that the number of people missed by the census could easily equal the estimated population loss in the First District as well as in Black-majority districts throughout the country. If the Census Bureau did not adjust its counts to add those missed, he wrote, such “mechanical” use of its population figures would continue to “mask the characteristically racist and fundamentally undemocratic manner in which the American political system has historically functioned with regards to Black people and other oppressed nationalities.”

Bailey had arrived at this conclusion using data made available by National Urban League research director Robert Hill, one of the people most responsible for shepherding the issue of the differential undercount into public view in the 1970s. The ICBS had invited Hill to present on a procedure he had developed, through which the bureau could do just that—eliminate the undercount of every racial and ethnic group through an adjustment process.

This procedure, what he called somewhat understatedly the “synthetic method,” was based on the widely accepted statistical assumption of the null hypothesis, that there is no significant difference between a population and smaller subpopulation groups. So if the bureau could calculate the national undercount for a certain demographic group, say young Black girls under five, or homeowning white men ages 35-39, then they could also calculate that rate of undercount (or overcount) at every area level from Illinois to Cook County to Chicago and assume it to be accurate. Or, more technically, not statistically different.

Crunching the undercount numbers for states and cities was something that the bureau had previously declined to do. But such numbers would be necessary for census data to be adjusted.

While the ICBS was eager to acknowledge the work of the bureau, and especially the regional Chicago office under director Stanley D. Moore (who was also present at the conference alongside regional coordinator for service programs Mary Grady, who had started a Black Census Day in Chicago), they were firm that there was no way forward without adjustment. “This is the era of Watergate, of COINTELPRO, and other gross government violations of the rights of people,” Alkalimat had written in a statement for the congressional oversight hearing in Chicago, just weeks before the conference. There was little to suggest that the Census Bureau could counter such levels of deep-seated government distrust, especially within communities of color.

As an institution, the bureau’s response to the suggestion of adjustment was mixed. Barabba persistently denied that adjustment would increase the accuracy of the count. But at the conference, some of his former and current staff spoke enthusiastically about the prospect. That included the bureau’s former acting director Philip Hauser—a respected University of Chicago sociologist who had previously gone on the record saying that the bureau was “dumping money down the sewer” and that taxpayers wouldn’t be out of line to tell census takers to “go to hell”—and the current deputy director, George Hall.

Despite the bureau’s ambivalence, it was clear that Hill’s method was not going to fly under the radar. By 1979, Stanford Research Institute had recommended its use in the Office of Revenue Sharing, and the bureau had used the method to calculate the effects of adjustment on funding and political representation in response to a demand for such information by the National Urban League in 1975. The bureau was planning its own undercount conference for February 1980, and had even invited Hill to present his work.

But, at least in the winter of 1979, it wasn’t yet convincing to the powers that were. When the ICBS had organized its conference, none of the major foundations Bailey or his cochairs had approached had agreed to provide any funding—which is why its member departments had covered the expenses. (“They love for us to talk about nothing,” Hill had complained at the conference, and one could basically see his eyes rolling through the text. “I find there is a correlation with lack of funds and seriousness,” he added. “When they saw the agenda that was here, they knew that these people were serious.”)

The lack of public support from traditional gatekeepers wasn’t anticipated by the ICBS, but it didn’t slow them down. Instead, they invested more to make sure the conference was recorded and printed to provide a record of the Black intellectual production of the time. “It is a mistake to think that the only pressure to be brought regarding the undercount issue is in that undercount meeting [hosted by the Census Bureau] in Washington in February,” said Bailey. “In fact, this conference represents pressure. Who demands an adjustment, the Census Bureau asks?”

Six months later, in May 1980, Detroit mayor Coleman Young filed a lawsuit against the bureau and its parent agency, the Department of Commerce, for their failure to count Detroit’s residents accurately and for the impact those omissions would have on city funding and on political representation for Black and Brown residents. Young demanded that the bureau adjust its data to mitigate the projected effects of the undercount in Michigan. His executive assistant Arletta Douglas had been in attendance at the ICBS conference the prior fall.

The count hadn’t been completed yet, but the bureau had shared preliminary numbers with Detroit along with other major cities as part of a new program designed to decrease the undercount. Young found that the bureau’s numbers departed significantly from the city’s calculation of its population.

The case, Young v. Klutznick, was closely followed by the media, but it wasn’t the only one that year. At least 51 other cases were filed against the bureau, the majority filed by cities, which had also come to question whether “Get Out the Count” initiatives would be enough to minimize the differential undercount.

In court, the Census Bureau took a hard line, arguing that adjustment wasn’t permitted by the Constitution, which required an “actual enumeration” of every resident. Young’s legal team, which was joined by Robert Hill and Philip Hauser, and a lawyer on behalf of the U.S. Conference of Mayors as amicus curiae, argued that not adjusting the data gave rise to a greater constitutional violation, that of the One Person One Vote rule.

The court ruled in favor of Young. In testimony, the bureau’s own officials had admitted that the census regularly makes millions of adjustments as part of “imputation,” the process of adding people who hadn’t responded to the census back into the count. In 1970, these imputations had totaled 4.9 million, just about half of all the people whom the Census Bureau hadn’t been able to reach.

“The decennial census undercount results in dilution of plaintiff’s vote relative to the votes of whites within the State of Michigan,” deciding judge Horace Gilmore concluded. “The undercount gives rise to a constitutional violation of the one-person, one-vote principle because blacks are simply not counted as much as whites, and this is particularly true in the inner cities of this country.” He ordered the bureau to come up with a plan to adjust its data—not just for Michigan, but for the whole country.

Gilmore’s order would be reversed in 1981, before the bureau could move forward with adjustment. In the Sixth Circuit Court of Appeals the judges ruled in favor of the bureau on a legal technicality: The financial and political harm Mayor Young had claimed in 1980 was hypothetical, based on “a state of affairs not yet in existence,” and therefore, Detroit didn’t have the standing to file suit in the first place.

In fact, most data from the 1980 Census had been released and analyzed by this point. The numbers showed only a slight decrease in the national Black undercount, from 7.7 percent to around 6 percent, as well as in the differential undercount between Black and white people, from 4.4 down to 3.7. But these were national averages, and belied the more severe undercount in hypersegregated cities like Detroit and Chicago, where Black and Brown neighborhoods like South Lawndale were undercounted by up to 8 percent, twice the city’s estimated undercount of 4 percent. But those rates may well have been higher: the Northeastern Illinois Planning Commission, which had estimated the undercount for the city and larger region in a 1989 paper, warned that their calculation of undercounts in Black neighborhoods were probably underestimates.

In the Sixth Circuit case, dissenting judge Damon Keith said that his colleagues were in the wrong, “avoid[ing] grappling with a serious question of constitutional harm.”

Credit: Amber Huff

Over the next six years, the bureau’s associate director for statistical methodology and research, Barbara Bailar, who had previously testified against adjustment in the Young v. Klutznick case, quietly oversaw the development of a procedure to adjust the final numbers. The work was cut out for her team: the number of demographic categories had ballooned from 96 in 1979 to 1,392 in 1990. That meant that there were more than ten times as many undercount rates to calculate and then redistribute in the process of adjustment. And while the process of distributing the undercount in large cities was relatively straightforward, doing so for counties and cities with smaller populations, more sensitive to statistical adjustment by nature, created a point of unease for some demographers.

To Kenneth Wachter, a University of California-Berkeley professor emeritus of statistics, adjustment represented a Pandora’s box of errors that could be introduced into the count. He worried that the post-census survey of 400,000 people was too small to be used as the basis for adjustment of the entire country’s population. There were many ways, he said, in which the bureau could err in matching or identifying that survey’s respondents while calculating the undercount. That would be all well and good in the process of assessing its own work—but in adjusting, a small error could multiply exponentially. To allay the concern, the Census Bureau had implemented several procedures, one of them “smoothing,” which removed outliers and more extreme data points before and after adjustment. But these added yet another set of technical issues, Wachter said.

“A lot of these formulas work well if the world behaves,” Wachter told me when we spoke by phone in November. “But in the context of the 1990 Census, on this huge scale, there were no good ways to check those assumptions. The real world is very messy.”

The field of demography and statistics was split by these questions. Wachter testified in several cases through the 1990s, joining the anti-adjustment side in arguing that the merits of adjustment were outweighed by the consequences of creating new and potentially flawed population totals for all the nation.

When I asked how he balanced the seriousness of the introduction of error against the need for equity in the count of Black people and people of color, Wachter reiterated what comes to sound like a mantra by the end of our phone call: “Places, not races. Shares, not counts.” Earlier, he had asked me to repeat this back to him.

What Wachter was talking about was “distributive accuracy,” which came to define the conversation around adjusting the 1990 Census. The ICBS and proponents of adjustment had been pushing for distributive accuracy by race, in part to end inequities in how money and political representation were distributed. But what Wachter was saying was that the conversation shouldn’t in fact be about race but about place—making sure that states and cities are receiving their “proper” share. Adjustment, he believed, would distort this distribution.

Certainly, this was not a point of universal agreement. Many scholars upheld the importance of distributive accuracy by race, and it was irrefutable that adjustment, in adding the disproportionately Black and Brown people who had been missed by the bureau back into the count, would increase the share of large, diverse cities—and thereby decrease the shares of states with whiter, less diverse populations. But it remained uncertain how much adjustment would rectify the undercount in the most undercounted areas in those big cities, like Chicago’s oft-cited Robert Taylor Homes in Bronzeville, which was undercounted at 29 percent in 1990 but only slated to see its population increased by a fifth of those who were missed.

Demographers and statisticians, sometimes from the same university department, could not agree on whether adjustment procedures would bring census numbers closer or further from the nation’s “real” population. To some, it started to feel like there was no good outcome. Leave the count unadjusted, and the differential undercount persists. Adjust the count, and potentially introduce other errors. As one New York Times journalist concluded somewhat pessimistically in 1991, “heads you lose, tails you lose.”

Nevertheless, by 1987, Bailar had expressed confidence not just in the processes they had fine-tuned but in the fact that adjustment would improve the overall quality of the census data. The post of director was empty at the time, but by the time Barbara Everitt Bryant took the position in 1989, she too expressed unequivocal support for adjusting the 1990 count.

But shortly thereafter, Secretary of Commerce Robert Mosbacher shut down the conversation, overriding the bureau’s tentative plan to adjust. Cities and states, still worried about the prospect of the undercount, once again filed suit. This time, certain states like Wisconsin also filed suit to block adjustment on the grounds of distributive accuracy, claiming that they stood to lose funding and political shares if their population data was adjusted. Both sides expressed outrage over what they perceived to be political interference.

In response to the Department of Commerce’s override, Bailar resigned from her post. “It was a political decision, and if they had said it was a political decision, I might have lived with it,” Bailar said in an interview with Education Week. “But they tried to make it look like a technical decision, and I couldn’t live with that.”

To Bailar, and to proponents of adjustment, the Department of Commerce was overstepping its bounds as a parent agency to the bureau. The major case filed against the Department of Commerce by New York argued that the decision was “arbitrary and capricious,” and beyond that, in violation of the Administrative Procedure Act, which regulates the actions of federal agencies.

Increasingly, the debate over adjustment was viewed as a Republican-Democratic standoff, with Democrats pushing for more money and representation for the cities they often represented, and Republicans fighting against losing the same for their states. “If you look at the people who are usually undercounted—blacks, Hispanics, Indians, the homeless, and people who live in urban centers—well, they are not the kind of people who are registered Republicans,” Bailar said to Education Week.

Though the partisan framing belied the more complex situation, there was certainly some truth to it: a report released by a Republican redistricting expert in 1994 predicted that the GOP stood to lose 24 seats if census data was adjusted. Whether the reverse was true was not as certain; a Brookings report argued that the political and financial benefits of adjustment were tangible, but minimal, pointing out that only several hundred million dollars of federal funds stood to be reallocated differently and that Republicans excelled at the art of gerrymandering, and the majority-minority district.

The judge on the New York case had ordered the Department of Commerce to conduct the post-census survey and convene a special advisory panel of experts before making a final decision regarding adjustment. The panel split four-four, with director Bryant in favor of adjustment and the under secretary of commerce for economic affairs against, though most agreed that the adjusted data set was more accurate than the original. Mosbacher acknowledged the majority consensus in 1991 when he announced for the second time that the census would not be adjusted, but insisted that “strict numerical accuracy”—the approximation of the true number of residents in the country—came second to distributive accuracy, which he said would only be improved in 29 states.

When the New York case was taken to the Supreme Court, the justices set what would become the standard for upholding adjustment-related decisions by the secretary of commerce: he had been neither “arbitrary nor capricious” in his decision not to adjust, and so the decision would stand. In 1999, when deciding another case filed by Indiana residents concerned about the dilution of their voting power, the Supreme Court then blocked the use of adjusted data for congressional apportionment—leaving it open, however, for the purposes of funding allocation and local redistricting.

To the statisticians who ran the bureau, the legal fight that began in 1980 had contributed to the politicization of adjustment and the takeover by the Department of Commerce. “The lawsuits have diminished the bureau’s autonomy,” Bryant complained, “moving adjustment decisions away from the purely statistical arena.”

Could the bureau have done more to make adjustment a reality? “This was the biggest statistical enterprise that had ever been conducted,” Wachter said. “It wasn’t that one side could prove they were worse, and another side could prove they were better. It couldn’t really be established whether they were better or worse.”

By many metrics, the accuracy of unadjusted census data has increased. Every year since 1990, the net and differential undercounts have decreased. In 2010, there was an overcount of 0.49 percent, down from 1.61 in 1990; the Black undercount had been more than halved from 1990 from 4.57 to 2.17 percent, the Hispanic undercount from 5 to 2.85 percent. The differential undercount between Black and white people remains, however, at around 3 percent. This steady decline in the overall undercount, Wachter said, is due to improved data and especially to the introduction of the five-year American Community Survey, which has given the bureau access to higher-quality data.

But charting census accuracy by undercount rate gives only a partial picture of progress toward racial equity. In a 2019 paper on the bureau’s data for omissions—that is, the full original undercount, the total number of people who don’t respond to the census, including those who the bureau “imputes” back into the count—demographer William O’Hare found that there remain dramatic racial disparities. Compared to the 3.8 average omissions rate for white people, Black people had an omissions rate of 9.3 percent, Latinos 7.7. The differential gap between Black people and white people remained low in terms of how many people the Census Bureau was adding or removing as they calculated who they had counted twice and missed. But what O’Hare found was that even as their average undercount had dipped to 2.17, Black people were still being disproportionately omitted from the count by 5.5 percent. He recommended his data be used by states, cities, and local organizations to target their outreach in future census years.

It will be difficult to argue that there are no issues requiring attention pertaining to the 2020 Census count. Actions taken by the Trump administration in attempts to change the nature of the count—by including a survey question about immigration status and by threatening to exclude undocumented immigrants—have likely contributed to lowered response rates in spite of being struck down by the courts, as have complications because of the pandemic. It is almost certain that the undercount will rise, and with it, the differential undercount. A University of California-Los Angeles study published in early 2020 suggested that, even controlling for a general decrease in census responses because of COVID-19, the gap was growing between white people and people of color.

When I spoke with Bailey, he said that the efforts to undercut the 2020 Census remind him of what happened in 1980. The Federation for American Immigration Reform, a conservative-interest group now designated a hate group by the Southern Poverty Law Center, had filed a suit in 1979 to remove undocumented immigrants from the count. But this census, Bailey saw a difference in how the powers that be responded to accusations of bias and inaccuracy. “That dedication to accuracy is not now the case,” he told me. He said that it felt like Republicans have decided “we have a pandemic that we know is hampering the count—there is 40 percent that still hasn’t been counted—and there’s no way we’ll be able to complete an accurate count.”

There hasn’t been much mention of adjustment for the 2020 data, though Bailey thinks that the prospect of an increasing undercount for the first time in several decades means the circumstances are right for a new conversation. And that there may be redistricting battles to watch in the lower courts as states begin the process of redrawing their districts.

He’s not alone. In October 2020, incoming American Statistical Association president and chief methodologist at the Urban Institute Robert Santos told WBEZ that he expects legislation demanding adjustment will be filed by advocacy groups after the data is filed with the White House. “This will be perhaps the worst census ever,” Santos said. He anticipated the undercount would hit cities like Chicago hardest, leaving middle-class white suburban communities with the best count and a disproportionate share in funding.

But since experts never arrived at a consensus about the merits of adjustment, any renewed push would probably fail in the courts, where the decision of the secretary of commerce has almost always been upheld as “reasonable,” write lawyers Molly Danahy and Danielle Lang in a paper for the University of Memphis Law Review. (A notable exception is the 2019 Supreme Court ruling that deemed then-secretary of commerce Wilbur Ross’s push to add an immigration question both arbitrary and capricious.) There is, however, a “narrow but important” exception originating in a Seventh Circuit decision. If plaintiffs can prove an undercount is the direct result of “intentional discrimination” under the Voting Rights Act, it could bring the possibility of adjusted census data back to the table.

Forty years later, the conference that the ICBS hosted at the University of Chicago remains an important record in the archive of Black Studies and in the history of the Census Bureau.

On our Zoom call, Alkalimat interrupted me to bring up the 1840 Census, which he sees as crucial for understanding the stakes of the census and its data in the United States, where the Black undercount had long been a matter of law, enacted for the political and financial benefit of white landowners. The results from that year were used to prove that free Black people “tended to be more insane” than enslaved Black people. “You can believe that has been attacked, that has been exposed, and the Bureau of the Census in the U.S. government has never said anything about that,” he said. “So that remains part of the historical record of the census without any repudiation. Even within our scientific record, there are these examples of racism that fabricate an image that’s incorrect scientifically.”

And it’s not just what’s done with the data, but what data is thought fit to be collected in the first place that’s a matter of politics. At the ICBS conference back in 1979, Waymon Watson, director of the South Shore Housing Center, identified the bureau’s failure to collect data about crucial housing issues, such as redlining, segregation, and absentee landlordism, as a failure to serve Black communities, which needed this information to organize against harmful and racist practices. As such, it was a double-edged sword. As long as “the Census data [was] blind at the neighborhood level,” he said, distrust of the institution among Black people would linger.

No one at the ICBS conference suggested that adjusting the census would fully resolve its bias or its use in supporting the racism of the status quo. But the ICBS had seen adjustment as only part of the battle. Their other demands included the training and hiring of Black demographers to participate in census analysis, and building a public-facing educational campaign to engage the broader public about the census, the resources that are tied to it, and who is impacted by incomplete data.

Alkalimat said that he sees the past year of uprisings as a “wake-up.” “It’s like—what! Somebody turned the light on in the room!” he said, laughing. He hopes that this new expression of anti-racism will reverberate within the scientific community, and that scholars will take on more responsibility for census education. The question of how to make the census important and accessible to people, and how to transmit that knowledge, was a central part of that 1979 conference. To that end, Alkalimat expressed relief and a sense of conviction that the ICBS had recorded the conference proceedings, establishing a record of the work of Black Studies in recentering the significance and failings of the census. Today, several copies of that orange-covered tome sit in stacks at the Woodson Regional Library, its 680 pages reflecting lifetimes of research, organization, and the fight against racism in all its forms.

“We really wanted to make sure that this conference was documented,” Alkalimat said. “Because a lot of things happen and they’re forgotten.” v

This story was completed with support from Reveal‘s Seeing 2020 Census Reporting Collaboration.