Thanks to the Oct. 13 Supreme Court decision, the data gathering phase of the decennial 2020 U.S. census ended at 4:59 a.m. last Friday. (The early-morning deadline was thanks to the Hawaii time zone.) The abrupt ending, which had changed no fewer than three times over the fall, meant that census workers were still fielding responses and queries late into the night.
When you put the shortened response work into the hopper with the COVID-19 pandemic and a reduced timeframe for data cleanup, the once-in-10-years 2020 census might be uniquely flawed. As the president-elect of the American Statistical Association recently stated bluntly: “I do not believe that a fair and accurate census can occur.”
On one hand, a sub-par census count is both good and bad news for Minnesota, which leads the nation in self-reporting. According to one analysis, the state is far more likely to keep its eight existing congressional seats. On the other hand, the 25% of people who did not self-respond to the census are not evenly distributed across the state, and a bad count will certainly affect those folks.
“The number one variable that would predict a low response rate is the predominance of renters in an area,” explained Andrew Virden, who is the state’s director of census operations and engagement. “[Anywhere with] a high percentage of renters is likely to have a lower response rate, as renters tend to be more mobile, younger, less affluent, lower educational attainment, a higher [percentage of] communities of color, and more likely to be immigrants. So for a lot of reasons, renters are a particular challenge.”
Virden’s office, housed in the Department of Administration, is charged with trying to coordinate state, local, tribal, and nonprofit groups. They worked for months to boost response rates and increase Minnesota’s self-reporting numbers. But the difficulty with reaching renters is a familiar story in urban politics, where most local political structures are set up to empower people who already own property.
Here’s the way it works for the U.S. census: For any address or group that does not self-report, the Census Bureau sends out canvassers to gather information. They do the best they can, sometimes using interviews with landlords, neighbors, or even postal workers. In each of these cases, creating the demographic estimates is a difficult process and often unreliable, and that’s the more reliable method of population guessing.
“The information is not as good when you rely on [proxies], and then one level down from the proxy info is the administrative records,” explained Virden. “The Feds might use something like IRS tax records to try to figure out what the count is at a given address, [and] that is the least reliable format.”
With more political and public health hurdles in place for the 2020 census, the end result is that Minnesota is all but assured to undercount its already marginalized communities. These groups include people of color, immigrants, renters, young people, and the state’s American Indian tribes. On the flip side, wealthy white people are more likely to be overcounted, particularly if they own a lake cabin or other second residence. One pre-COVID estimate suggested that, if it goes badly, the census might overestimate the state’s white population by more than 40,000 people.
Geography of reporting: a complex picture
With the data collection phase now complete, Census Bureau data already shows stark disparities on self-reporting rates. According to the map, the state’s lowest-reporting counties include Cass, Aiken, Cook, and Lake of the Woods, all sparcely populated areas of northern Minnesota. At first glance, Hennepin and Ramsey Counties do well by this measure, with self-reporting rates nearing 80%.
But when you zoom into tract-level numbers, the urban picture becomes far more complex. The areas with the lowest response rates include the census tracts and neighborhoods with high renter and POCI populations, mirroring the now-familiar map of metropolitan disparities.
Many of the 2020 census struggles can be chalked up to changes made by the Trump administration, including both a delay and cuts to census funding. But the sudden emergence of the COVID-19 pandemic in the United States added another level of chaos to the nascent count. In particular, Andrew Virden worries that the pandemic impacted counts of college-age adults who were displaced back in March.
“COVID really affects college students,” he explained, pointing to the overlapping timing of the COVID shutdowns and the census’ first round of mailers. “Undergrads at the University of Minnesota went on spring break, and many never returned to campus. In a city like Minneapolis, which has grown by 50,000 since the last census, not counting over 20,000 off-campus students is a big deal.”
Indeed, looking at the Census Bureau data, the University of Minnesota campus area jumps off the map with the lowest response rates in the city, which could make a dramatic difference in the demographic information. And a newly released map from the City of Minneapolis data dashboard, by contrast, points to the Whitter neighborhood as the one with the largest potential undercount. Whittier also has a very high proportion of renters.
“Minneapolis could effectively lose half of its actual population growth over the last decade,” Virden warned. “That’s going to affect Northfield, Bemidji, Duluth, Mankato, Winona, any city that has a campus. All of these cities are at risk of losing political power and political resources.”
When asked for comment on the potential undergrad undercount, the Census Bureau’s Michael Friedrich explained that the bureau tried to offset the undercount by producing a video and a .pdf fact-sheet aimed at college students. With so many students thrown into logistical chaos from switching to online classes and, in many cases, physically leaving their homes, it’s easy to imagine the census message getting drowned out. The low campus-area numbers suggest that the bureau’s outreach efforts did not bear much fruit.
Effects of miscounting
Undercounting a neighborhood or population is not just a matter of concern for wonky demographers. Any miscount has two key long-lasting effects: first, it reduces the amount of federal funding going to undercounted cities or counties; second, it affects redistricting at the state and local level.
In the first case, dramatic undercounts of renter populations in core cities or rural counties reduces overall funding for things like Medicaid, SNAP benefits, and more. In an era of increasing poverty and waning social service budgets, that difference can have a dramatic impact on vulnerable populations.
The second problem of undercounting comes in the form of redistricting — in other words, the literal shapes of city, county, and state representation. A widespread undercounting of thousands or tens of thousands of renters would reshape the local political landscape by overweighting wealthier, whiter homeowner neighborhoods areas at the expense of high-renter districts. In a city like Minneapolis, where the average City Council member represents just over 30,000 people, an undercount of renters would shift district-level politics, granting more seats to wealthier parts of the city for the next 10 years. The same kind of bias would hold for the Hennepin County Board and state Legislature seats, which in many cases already exist in a delicate balance between suburban, urban, and rural constituencies.
With responses officially halted, the census turns to the more arcane process of data cleanup, or as Virden describes it: “a very complicated process, the data cleanup, that is not as sexy.”
The cleanup phase involves things like “differential privacy” or ensuring that minority identities are masked — watch this video if you want to dive deep into the demographic weeds — and other ways to use other data to create the most accurate count. Typically, the cleanup takes about six months, but after the Trump administration order and the Supreme Court decision, this year’s process will end by Dec. 31. Last Friday, Commerce Secretary Wilbur Ross pledged to make the results public before submitting them to the Trump administration, though the precise details of what that means remain unclear.
“When the extension of the apportionment date was not granted, this is also problematic,” warned Virden. “It may ultimately affect the quality of the data that comes out, data used for things like apportionment, redistricting, and state funding formulas. That’s not something that [the State of] Minnesota has any influence over, but it’s very important that it is done right.”
When the counting is done and the numbers crunched, the political and viral assaults on the vaunted U.S. census will likely result in what one expert termed “one of the most flawed censuses in history.” The best-case scenario is that census workers doing the unsexy work of data collection and cleanup somehow deliver a quality count on short notice during a pandemic. But barring a numerological miracle, the worst case is that undercounting ensures that the state’s political landscape, already balanced against renters, immigrants, and younger Americans, remains uneven for 10 more years.