Why these Arizona towns are questioning their census counts and the revenue they direct

Geoff Hing
Arizona Republic

U.S. census undercounts – population numbers that fall below the actual population size – could mean less funding for transportation or nonprofits in Flagstaff, a college town with a substantial Native American community surrounded by lower-income rural areas.

In Superior, a mining town in the Tonto National Forest, they could make it difficult to build fire breaks to protect the town from fires like 2021’s Telegraph Fire.

In Somerton, an agricultural community near Arizona’s southwest border with Mexico, they could make it hard for the city to have enough police and fire staff to match its population.

Every 10 years, the U.S. Constitution requires a count of the U.S. population, and concurrently, the Census Bureau attempts to make sure those counts are accurate by conducting a Post-Enumeration Survey.

The counts and the accuracy of those counts are vitally important.

How census undercounts hurt cities

Not only are they used to determine the number of congressional representatives and for drawing political districts, but they guide more than $1 trillion in federal funding across hundreds of programs that impact even the smallest communities, according to funding analysis by George Washington University Research Professor Andrew Reamer.  

In Arizona, more than $29 billion in federal funding is directed by population data derived from the census. Like the nation as a whole, Medicare and Medicaid, which provide health coverage for older people and a variety of low-income groups, account for the most funds tied to census data, but the census also guides federal funds for roads, schools and food security.

Superior is one of the Arizona towns affected by 2020 census undercounts.

In addition to federal funds, census counts, and population estimates based on them, determine the share of four kinds of state tax revenue that counties and incorporated cities and towns receive. These revenue sources are particularly important for smaller communities. 

If cities and towns show a decline in population – even if that decline is inaccurate – it can mean hundreds of thousands of dollars in lost revenues.

For towns like Superior, an inaccurate count could mean the difference between being able to fund services or retain employees.

Superior is a community that hasn’t seen massive growth, but is still growing and needs to maintain aging infrastructure, said Superior Mayor Mila Besich. “We’re the towns that really can’t afford any loss to our budget, and our opportunities for other revenues and tax base are incredibly limited.” 

When initial estimates for shared revenue distributed based on the 2020 census were released, Superior was concerned that the loss in revenue could mean immediate tough staffing decisions and that the city wouldn't have money for an unexpected crisis. Updated figures made those concerns less acute because the overall tax revenue collected is higher, giving Superior a proportionally smaller share of a bigger pie. Still, "the reality is our budget would be much healthier and stable had we not lost population according to the census counts," Besich wrote in an email. "This has long-term repercussions."

Mila Besich, mayor of Superior, hiking in Arnett Canyon.

What data shows about census undercounts

The Arizona Republic reviewed census data to identify cities and towns in Arizona with a substantial portion of the population in groups that were undercounted by the 2020 census. It also studied places where there was a large difference in 2019 population estimates and 2020 census counts, or where the city was at the cusp of losing or gaining population. The Republic spoke with officials and other stakeholders in these places to understand their experience with the Census, the impact of the counts and whether they intended to challenge the results. 

From this reporting, The Republic found:

  • Across the country, Black people, Native Americans living on reservations, Hispanic or Latino people and young children were among groups undercounted by the 2020 census. The undercount of Hispanic people was greater in 2020 than in 2010 and estimated to be around 5%. In Arizona cities like Somerton, where the Hispanic population represents 97% of the total, the undercount could mean cuts to city services.
  • Other groups were not reflected in the Post-Enumeration Survey, like rural, lower-income communities and those with distrust of the government also were undercounted by the 2020 census. Colorado City, for example, a predominantly white town on Arizona’s northern border, had a substantial difference between population estimates and census counts. The town has a community that's uniquely distrustful of outside governments and that is recovering from the influence of a religious leader now serving life in prison for child sexual assault. 
  • Some city leaders said the transformation to an internet-first census in 2020 contributed to undercounts in areas that lacked internet access or where residents lacked internet skills. In Arizona as a whole, 52% of the population completed the census using the Internet. In many areas, Internet completion rates were much lower. 
  • The COVID-19 pandemic impeded census operations in ways that officials fear led to undercounts in other areas. The pandemic delayed in-person interviews of households, which particularly impacted agricultural communities and college towns. By the time census workers started knocking on doors in places like Somerton and Flagstaff, itinerant farm workers and college students had already moved on for the summer. 
  • Some cities and towns saw staffing issues with census workers because of COVID-19. Turnover of census workers impacted at least one city. The Census Bureau did not provide all the workers that were needed in another, and those who were provided were not always familiar with communities they were asked to survey. 
  • Cities and towns are permitted to challenge the undercounts. But the process requires expertise, time and money not available to every municipality. The factors that officials say led to an undercount of their communities may not meet the criteria for the Census Bureau to adjust counts. For some places, the potential payoff may not be worth it. 

On May 19, the U.S. Census Bureau released state-level results from its Post-Enumeration Survey, and the official measures of error rates did not show a significant undercount or overcount for Arizona as a whole.

According to the survey results, six states were undercounted: Arkansas, Florida, Illinois, Mississippi, Tennessee and Texas, while eight states were overcounted: Delaware, Hawaii, Massachusetts, Minnesota, New York, Ohio, Rhode Island and Utah.  

Two months earlier, in March, the Census Bureau released results from the same survey that showed, nationwide, some demographic groups were undercounted, while others were overcounted.

Due to the sample size of the Post-Enumeration Survey, the Census Bureau said it cannot estimate over or undercounts for areas within a state such as counties, cities or towns, nor can it provide results showing how different demographic groups within a state were counted. So The Republic used Census Bureau data sets to identify places in Arizona that may face miscounts.

One way the newspaper was able to find places where census accuracy is questionable was by comparing cities and towns that had large discrepancies between their 2019 population estimates and their 2020 census counts. Though population estimates are not an exact count of the population of a place, large mismatches between estimates and census self response rates can represent a red flag signaling ensus counts may be inaccurate. The Republic then looked at the relationship between these differences and the self-response rate for those places. Census data collected when people fill out the census form for their own household provides the most accurate information.

‘Your voice is not being heard’

According to the 2020 census, Somerton is the city or town with the highest percentage of Hispanic or Latino residents in Arizona.

The census also shows that the city experienced a slight drop in population from 2010. That means it will receive a smaller cut of shared revenue from the state.

But city officials are contesting the drop in population and requesting that the Census Bureau review the city’s counts. 

“When you have a 90% Hispanic population,” said Somerton City Manager Jerry Cabrera, “not being counted obviously means that your voice is not being heard.”

According to The Republic’s review, Somerton’s census counts were impacted by the same problems that affected other Arizona communities.

Fewer than 46% of households in Somerton responded to the census themselves and only about 42% responded using the Internet – well below the self-response rates for the nation as a whole and for Arizona. 

Ground operations also started too late to count many Somerton residents who didn't self-respond, according to Hector Tapia, the former economic development director for the city who led the gathering data effort for the city's challenge. Agricultural workers living in the community on April 1 left to work elsewhere before they could be counted by census workers.

More:This Latino city near Yuma grew, but the census says it shrank

“When they go into those neighborhoods … they may say, ‘Oh, there's a vacant unit.’ No, it's not vacant,” Tapia said. “They are going to work in a lot of places. That's one of the main issues that's happened, especially in Somerton.” 

Rural areas struggled to respond using the internet, and a preference toward limited government in Pinal County didn’t help to boost response rates, according to Pinal County Attorney Kent Volkmer. 

“I mean it is a somewhat invasive and sort of counterintuitive process,” said Volkmer, the point person for gathering information about Pinal County’s 2020 census counts and assessing whether it makes sense to challenge those counts. “If you really believe that small government’s better, then you have this giant arm of government who is trying to keep a, you know, a complete census of everybody in their community and where they live and what their phone numbers are and all of their information, I think there is a natural tendency to not want to comply.”

"And, you know, it was Reagan who, decades ago, said, ‘The most dangerous phrase in the English language is I'm from the government and I'm here to help.’ I think that that is a big reason why the census wasn't successful,” Volkmer said.

‘Undercount is an understatement’

In one northern Arizona town, wariness of governments and institutions seemed to play an especially big role in counting the community. 

Colorado City saw fewer than one in three households complete the census themselves. This is a smaller percentage than in Arizona as a whole and one of the lowest of any city or town in the state. 

The 2020 census counted nearly 2,500 people living in the town, but population estimates for the year before showed a population nearly twice as large. The percentage difference between Colorado City’s estimates and counts is larger than the gap for any other city or town in Arizona.

Shirlee Draper, director of operations for Cherish Families, a social service provider that services Colorado City and nearby communities, said the town didn’t lose population.

The city has a low housing vacancy and schools that are ”bursting at the seams,” she said, adding that “undercount is an understatement.” 

According to Draper, the reason for the undercount in Colorado City is years of distrust between the community and the state and federal government, and damage to the community caused by Warren Jeffs, the former leader of the Fundamentalist Church of Jesus Christ of Latter-day Saints. 

Jeffs is serving a life sentence in Texas after he was convicted of child sexual assault.  

The FLDS Church split from the Church of Jesus Christ of Latter-day Saints and continued to practice polygamy, even after the Mormon church ended the practice in the late 1800s. In 2016, a federal jury found that the town governments discriminated against residents who were not members of the FLDS church. 

Draper said that in past decades, the FLDS Church encouraged participation in the census and workers who went door-to-door were familiar faces. The 2020 census, however, came as the church was in disarray.

For subscribers:Arizona is diversifying. Here's what it could mean for politics

“The census arrived right around the time that there was no trust in anything,” Draper said.

The problem now is that low or otherwise inaccurate census data could cause social service providers who serve Colorado City to miss out on funding, and having resources for community services is particularly important because Warren Jeffs effectively dismantled many of the city’s social structures. 

But the community is fighting to recover.

“What’s come back are people who are really, really resilient but who have been left with shattered lives,” Draper said.

Fostering that resilience is time-sensitive, Draper said. Young people whose families experienced trauma while Jeffs controlled the FLDS Church are now reaching adulthood and at risk of self-medicating or repeating unhealthy parenting practices from their families.

“We can’t wait 20 years to decide to do it. By then we are chasing horses that have tried to escape,” Draper said.

Students pass through central campus at Northern Arizona University on the first day of snow on Oct. 12, 2021.

'All those apartments were empty'

In Flagstaff, city officials also worry that the community was shortchanged by the 2020 census. But in Flagstaff’s case, it was university students who officials fear were undercounted.

City of Flagstaff Comprehensive and Neighborhood Planning Manager Sara Dechter said that one in three adults living in Flagstaff is enrolled at Northern Arizona University. So, missing students in the census means missing a big part of the city’s population.

“I think the biggest issue with the timing and university students is simply the miscommunications and delay in operations that happened repeatedly between March and July 2020,” Dechter said.

Leading up to the start of the 2020 census, the city and county made plans with the Census Bureau to have on-the-ground census workers follow-up with those who had not yet completed the census in student-heavy areas before graduation in May. But as the pandemic worsened and the university switched to virtual classes, many students didn’t return to Flagstaff after spring break. And, there were no field operations until the end of May when many students had graduated or left for the summer.

The Census Bureau tried to address pandemic-induced migration of college students. In April 2020, the bureau told students that they should report where they lived before spring break, where they would have lived on April 1 if not for the pandemic, not where they were actually living on April 1. But  this contradicted the pre-pandemic messaging that said they should report where they were living on April 1.

Flagstaff officials said many NAU students were not counted in the 2020 census. The students make up a significant portion of the city's population.

“It was very hard to make that pivot and I don’t think it was very effective,” Dechter said.

Census results for the census tract that includes both the university and some off-campus apartment buildings that cater to students had 25% of housing units vacant, Dechter said.

“What that tells me is they went when all those apartments were empty and they couldn't figure out who lived there,” Dechter said.

In addition to timing of on-the-ground census operations, some places faced difficulties with staffing and coordination.

Besich, the mayor of Superior, said the town had planned to hold an ice cream social and events at the senior center to encourage census participation, but these plans were cut short by the pandemic.

“COVID obviously limited any type of group activity which we were relying on to help people feel more comfortable with the census,” Besich said.

She added that the town cycled through three different Census Bureau contacts who were supposed to help the town access resources and manage census outreach, making it feel as though the town was an “afterthought.” Besich said that while neighboring counties took a more collaborative approach to census outreach, it felt like Superior was on its own to get a complete count.

Pinal County Attorney Kent Volkmer speaks to the media outside the Pinal County Superior Court in Florence on March 29, 2019.

Volkmer said Pinal County didn’t get its allotted number of field enumerators, the census workers who knock on doors to survey residents. Moreover, the enumerators that were assigned came from Tucson and weren’t familiar with communities in the county.

“It's not like we had locals,” Volkmer said. “We had complete strangers, in the middle of COVID, undermanned, knocking on doors.”

Enumerators from another area may not be familiar with smaller villages on tribal lands in the county. The Census Bureau found that, nationwide, Native Americans living on reservations were a significantly undercounted group.

In Somerton, the city had to hire its own workers, through a state job-placement agency, to be census takers.

“The Census Bureau didn't do it,” said Tapia, former economic development director for the city. “They didn't have the staff. They didn't have people, especially here in Yuma County. So the city agreed to pay.”

‘Reality, truth, none of that actually matters’

At an April 19 meeting, the Somerton City Council approved a final summary report and challenge that the city submitted to the Census Bureau. This challenge is through a Census Bureau program called Count Question Resolution.

The city’s challenge hinges on their analysis of building permits that show completed homes since the 2010 census. The 2020 census counted 4,149 housing units in the city, while the city estimates there are 4,891 units.

City Manager Cabrera said there are other signs that the population is growing, even as that’s not reflected in the census numbers. A new elementary school is opening in August and a high school is slated to open in coming years. The city is also seeing more small businesses and retail open.

Tapia, the city’s economic development director, said these discrepancies led him to dig into the counts. “Even though we have 800 plus new homes, they gave us only 97 units for credit. So when I saw those numbers  … that's when I got involved,” he said.

Tapia looked at satellite images from 2010, and more recent ones in a Census Bureau system to show new homes that have been built over the decade. The city identified about 300 housing units missing from the census counts.

“So it's off. There were blocks that were missing (housing units) from the Census Bureau, so that's very important for us, because that's one of the determinations for the appeal is if blocks were missing, and obviously they were,” Cabrera said. “They didn't show up on their system.”

The Count Question Resolution process only allows updates to data collected, but incorrectly processed, during the 2020 census. It does not allow for collection of additional data. A bureau explanation of the process said, “this is not an opportunity to redo the census or to conduct additional data collection.” While the process allows tribal, state, county and local governments to get more accurate counts for funding requests and planning, and updated counts are incorporated into Census Bureau population estimates, the process does not change the numbers reported from the 2020 census, including apportionment counts and data used for redistricting.

A successful challenge requires more than noting discrepancies in the count of homes. The Census Bureau will only update counts when it has discovered it has missed housing it knew about in 2020.

The Census Bureau’s guide for the program states, “Coverage corrections are limited to census processing errors, i.e., erroneous exclusions of housing identified as existing in census records as of April 1, 2020.”

These parameters don’t allow corrections for issues that arise during enumeration, like an undercount of households that didn’t respond to the census themselves and had to have their information inferred by a neighbor, administrative records or estimated statistically based on the information known about nearby households. The limited nature of the kinds of cases that can be brought to the bureau is frustrating to those like Pinal County’s Volkmer who believe there were undercounts for other reasons.

“There are some housing count cases, but essentially what they say is this is the way that we're counting. And we don't care if you don't think we count correctly. We don't care if you don't think that's a fair way to count. This is how we're counting and under our rules, if we miscounted, then you can bring it to us,” Volkmer said. “I somewhat jokingly say, you know the facts, the reality, truth, none of that actually matters.” 

In a November 2021 letter to the U.S. Census Bureau, Flagstaff officials said communication problems and the timing of 2020 census operations lead to difficulties counting the city’s population of college students.

The city asked the bureau to allow resubmitting administrative records that could aid in counting college students as part of the Count Question Resolution process. The city also asked that off-campus housing for students be considered part of the group quarters counts so that they could be included in updated counts.

“Part of what's the issue with the CQR is it doesn't cover things like this,” said Dechter, Flagstaff’s comprehensive planning manager. “Like some of it doesn't allow us to appeal this kind of finding so we have to use the criteria that they make available so we could have an undercount and have no way to question it or get a second look at it unless there was some kind of legislative action.”

For subscribers:Census undercount may have cost Arizona a congressional seat

At the end of May, the Census Bureau announced a new program called Post-Census Group Quarters Review, that aims to address issues with the count of people living in group quarters like university dorms, prisons and jails and nursing homes. However, Dechter said the city is still reviewing this new program.  She said the city needs to verify counts for privately owned, off-campus student housing that the post office may even consider as dormitories for mail delivery purposes. She hopes the city can submit these counts as part of the new program. The Census Bureau will accept cases through June 30, 2023.

Even though the group quarters review program has just been announced, Dechter said Flagstaff is finalizing a case under the Count Question Resolution program by looking at housing units that may have been missed during the census. Looking into discrepancies in the counts takes resources, Dechter said, but finding a few hundred households that weren’t counted could mean hundreds of thousands of dollars worth of funding for the city.

The Count Question Resolution process following the 2010 census did not result in any corrected counts for Arizona.

Nationwide, an analysis by the Georgetown School of Law’s Center on Poverty and Inequality Economic Security and Opportunity Initiative found that challenges after the 2010 census resulted in a net increase of only 527 people and 224 housing units across the entire country.

“So I can tell you when these numbers first came out 6 to 8 months ago, I would have told you I was confident that the county was going to mount a full-scale challenge. I do not know that that is the case as I sit here today,” Volkmer said.

Jurisdictions can also conduct a special census in the middle of the decade. However, Volkmer said, since jurisdictions must pay for this out of their own pockets, if the special census produces counts under expectations, it can mean the jurisdiction loses money.

Dechter said that many jurisdictions might not have the time or resources to pursue a challenge.

In Flagstaff, the city hired someone to go through hundreds of census blocks where there are signs that the count might be off to determine if there’s an explanation for any discrepancies, or if it represents housing units that were missed by the census. More than a pure geographic information system exercise, this verification takes a lot of time, effort and knowledge, Dechter said.

“There’s a lot of detective work,” she said.