Donald Trump’s most consequential legacy might be the debate he spurred about whether our democracy truly includes everyone, a choice evident in the 2020 election and Ketanji Brown Jackson’s Supreme Court nomination. Should public life in the United States be based on inclusiveness or on the hierarchy that prevailed for much of our history? That’s also the issue in the Trump administration’s hobbling of the 2020 census and how it distorted the current apportionment for the House of Representatives.
The large-scale errors in the census cost New York, Texas, Florida, Arizona, California, and New Jersey one seat each, and resulted in an extra representative for Minnesota, Pennsylvania, Oregon, Montana, Wisconsin, and Indiana.
You may recall then Commerce Secretary Wilbur Ross’s ham-handed scheme to include a question on citizenship status in the 2020 census. It was a ploy to depress minority participation, and it was batted down by the Supreme Court. But the bigger scandal was the census’s persistent funding shortfalls, understaffing, and truncated schedule. The result was the most error-riddled count in decades. The undercounting of Blacks and Hispanics and double counting of whites and Asians altered the allocation of congressional seats for the next decade.
Those wide-ranging errors are matters of public record, because the professionals at the Census Bureau obligingly report the decennial census undercount and overcount rates by race and ethnicity. Compared to 2010, undercounts in 2020 jumped from 2.06 to 3.3 percent for Blacks, from 1.54 to 4.99 percent for Hispanics, and from 0.15 to 0.91 percent for Native Americans on reservations and Alaskan Natives. Overcounts also shot up, increasing from 0.83 to 1.64 percent for whites and from virtually zero to 2.62 percent for Asians.
These substantial errors skew the apportionment of seats in the House because the racial and ethnic makeups of the states vary so widely. The share of Black residents by state in 2019 ranged from less than 2 percent in Montana, New Hampshire, New Mexico, and nine other states to more than 30 percent in Maryland, Louisiana, Mississippi, and Georgia. Similarly, Hispanics accounted for less than 3 percent of those living in Maine, Mississippi, and two other states, compared to between 27 and 50 percent of residents in Florida, Arizona, California, Nevada, and New Mexico.
In much the same way, overcounts tend to be concentrated among two groups—non-Hispanic whites and Asians. The share of white residents by state ranged in 2019 from 36 percent in California and New Mexico to 92 percent in Vermont and West Virginia. Similarly, Asians comprise less than 2 percent of the population of Mississippi, Wyoming, West Virginia, and 13 other states, compared to 9 percent in New York and Nevada, 10 percent in New Jersey, 15 percent in California, and 39 percent in Hawaii.
By applying the 2020 error rates to each state’s racial and ethnic makeup, we find that undercounts in the 2020 census deprived six states of a congressional seat; correspondingly, overcounts of white and Asian residents enabled six other states to gain one seat more than their populations warranted.
The debasement of the 2020 census did not have clear partisan effects. The more diverse states that lost out—states in which Black and Hispanic people account for between 33 percent and 52 percent of the population—include not only blue New York, California, and New Jersey but also red Texas and Florida and purple Arizona. Similarly, the unwitting winners include not only red Montana and Indiana but also blue Minnesota and Oregon and purple Pennsylvania and Wisconsin—all states with populations that are 80 to 87 percent white and Asian.
The effort by Trump officials to hobble the census did not involve overt manipulation of the results. Instead, they simply withheld the means to avoid serious errors. Here is how it happened.
Census forms are sent to addresses without the names of those who live there, and undercounts occur when those people don’t respond by mail or online or to census workers who visit non-responding addresses. Black and Hispanic people have been more likely to be undercounted for several reasons. First, errors are more common among renters, and Blacks and Hispanics have much lower homeownership rates than whites and Asians. Undercounts are also more common among lower-income people and immigrants, who are disproportionately Black and Hispanic, because they are more likely to fear that the Census Bureau will pass their information on to other government agencies. Even though federal law prohibits such sharing, most people don’t know that. In 2000, when I oversaw the Census Bureau as undersecretary of commerce, and again in 2010 after I’d left, the bureau spent millions of dollars on advertising that urged recipients not to be afraid. No such advertising aired in 2020.
Overcounts usually occur when people with two homes fill out census forms at both addresses and when college students respond from their college addresses and their parents also list them at their home addresses. These overcounts have tended to skew toward whites and Asians because their homeownership and college enrollment rates are substantially higher than those of Blacks and Hispanics. Targeted advertising can also limit the extent of these overcounts, but, again, no such messaging happened in 2020.
Other factors spurred the sharp increases in errors in 2020. For several years, Trump’s budgets denied the Census Bureau the resources to compile the tens of millions of addresses and test the information technology used by census workers. Perhaps most important, Trump officials prematurely ended the census ground operations that usually ensure more accurate results by visiting every non-responding or questionable address up to three times.
We can also pinpoint precisely how the outsized errors in 2020 changed the composition of the House by applying the complex formula the Census Bureau uses to determine the apportionment of Congress and by analyzing the census data on the numbers of people or “priority values” that merit each seat by each state. For example, New York lost one seat in the 2020 apportionment that we can be confident it should have retained: The census priority values show that New York would have held on to that seat if its official population had been just 89 persons larger—and the error data show that, based on race and ethnicity, New York had an estimated net undercount of some 61,750 people.
Those priority values also show that the official population in Texas—unadjusted for errors—was 3,100 people short of the Lone Star State gaining another congressional seat. Its error rates produced a net undercount estimated at nearly 464,500 people. Similarly, census data indicated that Florida fell 4,200 people short of receiving another seat in Congress, while its estimated net undercount totaled 192,500.
It’s the same story in three other states: Arizona was 6,600 people short of gaining another seat and its net undercount was 69,500. California, with a net undercount of 469,000 people, had an official population that was only 7,300 people shy of receiving another House seat. And New Jersey, officially 17,000 people short of gaining another seat, had a net undercount of 29,500.
Correspondingly, six other states had net overcounts that were much larger than the margins that entitled them to another seat in Congress—again, Indiana, Minnesota, Montana, Oregon, Pennsylvania, and Wisconsin. For example, with an estimated net overcount of nearly 50,500 people, Minnesota managed to keep its eighth seat in Congress by an official margin of fewer than 100. Similarly, Pennsylvania’s net overcount of more than 78,000 people dwarfed the official census margin of around 26,000, enabling the Keystone State to keep its 17 House members.
Officials in the Trump administration may or may not have conspired to use the 2020 census to shift seats in Congress based on the states’ racial and ethnic makeup, but that was the inescapable effect of their debasement of its operations. The outcome is also consistent with the GOP’s implicit (and sometimes explicit) consensus that power and legitimacy in America are inseparable from race and ethnicity. That view is simply incompatible with democracy.