99% Completion Rate Does Not Equal a ‘Complete and Accurate’ Census

As the U.S. Census Bureau nears the end of data-collection operations under a compressed timeline, the Trump administration is claiming that the government is on track to produce a “complete and accurate census.” The Census Bureau is releasing “total response rates” by state, which officials say is evidence that they will “complete” the enumeration of at least 99 percent of all homes. However, it’s what they aren’t telling us that matters. 

The total response rates only tell us how close the Census Bureau is to finishing data collection for known housing units. These indicators don’t tell us about the quality or accuracy of the count, such as whether all people in each household were counted, whether households were counted in-person, if the bureau will use administrative records to populate the home after one unsuccessful contact attempt or whether units were deemed vacant even if people were living there on April 1. 

First, households that don’t respond on their own are more likely to be reluctant to share information with the government out of fear that information would be shared with a landlord, law enforcement, or immigration officials.

Second, it’s harder to determine who was a household member as of April 1, the reference day for the population count. The delay in conducting this operation in 2020 due to the pandemic made this recall bias an even greater stumbling block to collecting accurate information.

Third, census takers who are unable to interview a nonresponding household member will try to find a neighbor or landlord to provide basic information about the residents. In 2010, proxy interviews made up 22 percent of all “completed” NRFU cases, and almost a quarter of those interviews did not yield useful information.

Fourth, the Census Bureau estimated that it could “enumerate” at least 6.2 million nonresponding households using administrative data from other federal agencies after one unsuccessful visit from a census taker. Experts believe that these records systematically leave out population groups, such as children and noncitizens, who already are at risk of being missed.

Finally, some housing units are vacant now but may have been occupied on April 1. The bureau also has to count people experiencing homelessness; people who live in group facilities, such as college dorms, prisons, nursing homes, and military barracks; and people living in transitory locations like RV parks, motels, and marinas.

Census takers can usually resolve the easier cases in their workloads quickly, which account for initial rapid increases in NRFU completion rates. In historically undercounted communities, however, where self-response rates have been lowest and the caseloads for in-person visits are highest, completing the enumeration accurately takes time. When the administration essentially shortened the timeline for in-person visits, the bureau streamlined some operations to finish faster.

Equally worrisome, nonpartisan organizations promoting census participation heard reports from enumerators who were instructed to skip questions and reduce the number of visits in order to finish sooner. These reports suggest that extending the follow-up operation through October, as expert bureau staff had recommended, would help ensure a better count. 

Indian Country poignantly illustrates the need for those extensions. American Indians and Alaska Natives have been ravaged by the pandemic. Since the bureau’s April 13 announcement, dozens of Native villages and Indian reservations have been closed to mitigate the virus’s impact and to protect the health and safety of vulnerable populations. Update Leave operations, in which census workers drop off packets to households lacking traditional mailing addresses, did not occur in much of Indian Country, leaving tens of thousands of American Indians without information needed to complete the census.  

The impact of pandemic-related closures is evident in the exceptionally low self-response rates throughout Indian Country. As of Sept. 16, among the reservations in the Lower Forty-Eight for which self-response rates are reported, only 5 percent of reservations are at or above the national rate of 66 percent. Nearly 62 percent of all reservations are more than 20 percent below the national self-response rate. The Navajo Nation, which comprises approximately 200,000 people living in Tribal Lands in Arizona, New Mexico and Utah, is at just 19.7 percent. The Census Bureau has only had staffing levels in the Navajo Nation of about 40 percent of the planned enumerators, which is insufficient to complete the count in 10 days.  

Tribes have spent months coordinating with the Census Bureau and preparing their campaigns around the Oct. 31, 2020, deadline. The efforts to complete the count have only begun recently on the reservations, reflected in the improving response rates. Now is not the time to cut short the census to the exclusion of America’s First Peoples, who suffered the highest undercount of any population group in the 2010 census.  

This is not a normal census. Operations have been delayed considerably due to COVID-19 and are now further complicated by hurricanes and wildfires. This year’s operations are more likely to miss people, count people twice, or fail to collect accurate information about critical characteristics such as race and age. Expert staff also noted that they needed six months for quality assurance and data processing activities in the wake of these challenges. That period has been cut to three months.

The proportion of households that haven’t been counted varies by state. Measuring success by the total response rate or pushing to reach the 99 percent goal everywhere by Sept. 30 will make it harder to collect accurate data and force the bureau to fill-in information for more households using less reliable sources. 

A rushed census won’t be accurate. Every state is facing the likelihood of an undercount. Congress must push back the statutory reporting deadlines for apportionment and redistricting data by four months each, as the administration itself requested in April, before it abandoned its support for these extensions without explanation or warning. Lawmakers must extend the deadline and give the Census Bureau the time it needs to finish counting operations and conduct vital, complex data processing and quality assurance activities thoroughly and carefully. 


Dr. James Thomas Tucker is pro bono voting rights counsel to the Native American Rights Fund and an attorney with the law firm of Wilson Elser Moskowitz Edelman & Dicker LLP in Las Vegas; he serves as vice chair of the Census National Advisory Committee.

Morning Consult welcomes op-ed submissions on policy, politics and business strategy in our coverage areas. Updated submission guidelines can be found here.

Morning Consult