The Census Bureau announced Thursday that it will not produce its annual American Community Survey, which provides detailed demographic data widely used for research and billions of dollars in federal funding decisions, because of how the coronavirus pandemic skewed survey results.
The data for the 2020 ACS did not meet agency standards, Census officials said. Instead, the agency will release an “experimental” dataset that includes less information than the normal release.
As a result of the pandemic, ACS responses skewed toward wealthier, more educated individuals — too many to account for with statistical adjustments, Donna Daily, chief of the bureau’s ACS office, told reporters. The Census Bureau could not move ACS data gathering from the height of the pandemic lockdowns last spring, unlike the delay the agency used for the decennial census.
“We know people rely on [American Community Surveys] and make decisions and plan for their communities. It's critical that estimates truly represent their communities, and standard 2020 ACS one-year data products would not be able to do that,” Daily said.
Census publications have cited ACS data as influencing more than $600 billion in federal spending each year. However it’s not certain yet how much the Census Bureau’s pullback of ACS results will impact federal programs.
Daily said the agency has started working with other federal agencies to help them shift their plans following the problems with ACS data. She also said the agency would not produce the annual coverage measure of health insurance in the United States as a result of the disruption.
“We are here to help and provide tools for them to determine the best way for them to make their decisions in their agencies using the data that we can provide to them,” she said.
Unlike the decennial census, the ACS is a sample survey that relies on data from a smaller number of people to represent the larger population and is released annually. The ACS normally provides data drilled down to geographic areas with as few as 65,000 people. Census officials said that this year some data tables would be available at the national and state levels, with the possibility of data available at lower levels of geography, but they did not specify which ones.
ACS tables typically include a wide range of demographic, social and economic information, ranging from language proficiency and internet access, to income, education and even commuting times.
The Department of Education’s Title I funding program, which provided $16 billion in grants this fiscal year to schools with large low-income populations, relies on an annual poverty estimate the Census Bureau produces based on ACS data. The House is expected to vote this week on a funding bill that would increase the program to $36 billion.
David Waddington, chief of the agency’s decennial statistical studies division, said the Census Bureau is still evaluating how the pandemic affected the agency’s supplemental estimates, including the one used by the Title I program.
Other federal measures, such as the Department of Housing and Urban Development's small area fair market rent calculation, explicitly mention one-year ACS data in their enacting regulations.
HUD’s Community Development Block Grant program refers to the ACS more generally in the regulations concerning funding decisions. That program distributed more than $3 billion in grants this fiscal year.
Some programs that use census data, such as the Small Business Administration’s HUBZones, have paused new calculations until after the release of the 2020 census results.
Daily said the agency plans to produce a more detailed report in October on the data problems and efforts to address them. The agency plans to release the experimental dataset in November, along with information about the statistical adjustments it used.
The agency created the ACS in 2010 as a replacement for the long-form census questionnaire that used to be sent to a subset of American households. The survey asks about 3 million households a year more extensive questions than the 10-question decennial census, rolling results into one- and five-year data releases.
The pandemic could still hurt the five-year dataset, as the agency continues to work through results, according to Mark Asiala, an assistant division chief in the Census Bureau's decennial statistics studies office. The agency received about two-thirds of the normal number of responses to the survey in 2020, Asiala said.
He noted the most troublesome portion represented about six months out of a five-year survey period. However, that data will remain part of ACS datasets the agency produces until the mid 2020s, as each five-year dataset will include the 2020 survey through 2024.
In addition to direct federal spending decisions, ACS data undergirds planning decisions for roads and bridges, schools and more.
The Current Population Survey, which the Census Bureau collects along with the Bureau of Labor Statistics, ran into similar problems in the pandemic, according to agency publications. Working papers published over the past year have laid out problems measuring individual metrics due to the pandemic, such as changes in health insurance, but Thursday’s announcement is the largest change yet.