Where a Census Undercount Will Hurt (or Help) Most

On Wednesday, Nebraska Governor Pete Ricketts vetoed a bill that would have authorized a Complete Count Committee tasked with preparing the state for the 2020 Census. While the state legislature passed the bill by a veto-proof majority, lawmakers adjourned for this legislative session last Friday.

An undercount in the 2020 Census could cost Nebraska dearly. Every person in Nebraska who fails to be counted over the next decade will cost the Cornhusker State $21,000 in federal funds over the next decade, according to a study from the University of Nebraska. A statewide undercount of one-tenth of one percent (0.1 percent) would cost $400 million.

But an undercount would not affect all Nebraskans equally. Under a low-risk scenario, Nebraska’s surging Hispanic population could be miscounted by as much as 1.98 percent (or 4,500 people), according to new estimates from the Urban Institute. And under a high-risk scenario—say, one in which the state actively avoids preparing for a census that includes a citizenship question—that miscount estimate rises to 3.61 percent (8,200 people).

Under no scenario does the potential for miscounting the state’s white population rise to one-tenth of one percent. Even a high-risk scenario figures a miscount of just 0.05 percent of white, non-Hispanic Nebraskans (800 people). In fact, the low-risk scenario predicts an overcount of 11,300 whites—nearly a full percent.

Nor would an undercount negatively impact every part of the state the same way. Nebraska’s Hispanic population doubled from 2000 to 2017, and many residents are clustered in cities like Lincoln and Omaha, where about a quarter of the population also consists of historically hard-to-count underage residents. Those cities would be hard hit by a miscount, losing funding and political power. Rural areas have seen a big increase in their Latinx populations, too: Rural Nebraska counties with populations today smaller than they were in 1890 have seen gains due to immigration, but those gains would be undermined by a census undercount.

And Nebraska is just one state! Across the country, states with the highest risk of an undercount include California, Texas, New Mexico, Nevada, Georgia, New York, and Florida, according to the new Urban Institute assessment. The report shows the ways in which demographic changes, new census-taking technology, and a  citizenship question that has been calculated to inflict partisan damage—plus old-fashioned heel-dragging in red states—could favors whites and Republican-leaning districts nationwide.

Some states stand to gain disproportionately from the decennial count. In a low-risk scenario—a 2020 Census in which things go relatively smoothly—states with large shares of white residents will see their populations overcounted. Idaho, North and South Dakota, Wyoming, Iowa, and West Virginia are among those that may get a disproportionate boost in congressional power and federal funds. White homeowners were overcounted in the last census: They are more likely to be counted at more than one address, for example, especially if they own more than one home.

That’s the very outcome dreamed up by the “Michelangelo of gerrymandering,” the late Thomas Hofeller, a GOP fixer known for his work in drawing up partisan districting maps. In a striking story, The New York Times recently reported that, after his death, Hofeller’s estranged daughter discovered a cache of documents that pointed to her father as the architect of the citizenship question, providing the language and logic for a tool that designed to give an edge in representation to whites and Republicans.

That campaign to rig the census for the benefit of one party now seems ever more likely to succeed, thanks to demographic changes, untested processes, and a refusal in a few states to abide by best practices recommended by the U.S. Census Bureau—the ever-daunting citizenship question notwithstanding.  

On the same day that Ricketts vetoed a bipartisan, conventional, Census Bureau–recommended Complete Count Committee for the whole state of Nebraska, a federal judge addressed for the first time the explosive new evidence of the alleged GOP conspiracy to boost white representation in the U.S. by adding a citizenship question to the 2020 Census. In a brief hearing, Judge Jesse Furman of the U.S. District Court for the Southern District of New York opted for a take-it-slow approach, scheduling future briefings in which plaintiffs will present evidence for what he described as “serious” allegations in the case.

The anticlimactic hearing was the tonal opposite of the motion submitted by plaintiffs outlining the “damning new evidence [that] reveals hyper-partisan and racially discriminatory motives at the root of the citizenship question.” It’s unclear yet whether the U.S. Supreme Court will weigh this new evidence when it issues its decision about the citizenship question.

The decision may determine the scale of any undercount in the next census, but at this stage that undercount may be a given. “An undercount in the 2020 Census is likely inevitable,” the report by Urban Institute senior research associate Diana Elliott reads. “The big question is by how much.”

With the next census only a year away, states that dread the prospect of an undercount are taking pains to try to avoid it. Stacey Abrams, the former Democratic gubernatorial candidate in Georgia, launched a new nonprofit, Fair Count, to complement her efforts to expand voting access across the state. Georgia is another state with a lot to lose: In a high-risk scenario, the state’s black population could see an undercount of 3.80 percent (136,600 people). Los Angeles launched its census-planning operations in 2017 to get a three-year head start.

The Urban Institute report points to a number of factors (beyond the citizenship question) that could deliver an undercount. One is low and unpredictable funding. Despite population growth and the new technology to be implemented for the first-ever digital census, Congress has apportioned less funding for the 2020 Census, and at a slower rate. Two of three planned dress rehearsals were canceled in 2018, leaving only the end-to-end test run in Providence, Rhode Island, to assess some 50 new IT systems.

Internet self-response will be another likely culprit in any miscount. For the first time in the history of the census, households will be able to opt to complete the questionnaire online. Census research shows that “advantaged homeowners” are most likely to self-respond online in the next census. But many households—and especially those that are already historically hard to count—lack reliable access to the internet. While an online census ought to be cheaper (and more accurate) for the Census Bureau to conduct, the Urban Institute notes that self response rates to federal surveys have declined over the years. And Reuters reports that federal officials worry that the 2020 Census could be subject to disinformation campaigns (“fake news”) via Google, Facebook, and Twitter.

Concerns about using administrative data records for the first time to improve counts seem picayune next to the threat that Russian troll farms could undermine the nation’s confidence in the census—but that, too, is another innovation that census monitors worry has been insufficiently tested. Administrative data favor older homeowners, for whom reliable data exist, over younger renters and children, for whom records may be spotty.

There are reasons to worry that this census could feature a dramatic undercount even beyond the alarming evidence that may or may not reach the Supreme Court’s bench. But experts nevertheless say that there’s a chance to right the ship.

“There is still time to ensure that representation in the 2020 Census is fair and accurate,” the report concludes. “By investing in outreach and engaging communities in a culturally sensitive way, there is time to ensure that representation will be better.”