The United States Census Bureau will deliver new population estimates for 2020 in November 2021. But it has cautioned people against using them, while trying to inspire confidence in a different headcount that itself deserves closer scrutiny. Trillions of dollars are at stake over the next decade, as well as control of Congress and the Electoral College.
By way of background, the Census Bureau provides two main population counts for the United States. They include annual population estimates based on sample surveys, and a complete “census” count once every 10 years. The annual estimates come from the American Community Survey (ACS) program, which is based on sample surveys covering a fraction of households. The once-a-decade (“decennial”) complete census totals are a much more extensive effort, theoretically covering everyone in America.
The decennial census results are used to determine apportionment – the number of seats a state holds in the House of Representatives. So the decennial census effectively determines the number of electoral votes for each state since each state has as many ‘electors’ in the Electoral College as it has Representatives and Senators in the United States Congress. The decennial census totals matter for other critical policies, including the distribution of hundreds of billions of dollars of federal funding for state and local governments annually.
In other words, those decennial census numbers are pretty darn important. But the annual population estimates matter, too, given that many government decisions are based on more timely population information.
The legal foundations for the decennial census were first laid in Article I, Section 2 of the U.S. Constitution, which called for an “Enumeration” of “the whole Number of free persons … and excluding Indians not taxed, three fifths of all other Persons.” The Enumeration was designed to provide the means for determining the number of Representatives for each state. The “three-fifths” calculation no longer applied after the Civil War and the end of slavery.
The annual ACS survey results are based on samples, not complete totals, and therefore are subject to sampling error. Describing the interpretation of annual survey results and census totals, the Census Bureau has stated that “Differences between the estimates and census counts are interpreted as error in the estimates and not the census counts.” This assumption may include or imply misplaced confidence in the decennial census totals, however.
In 2020, the ACS was only able to gather two-thirds of the responses it normally collects in the annual survey process. In late July 2021, the Census Bureau announced it would not release the annual ACS estimates for 2020, given the impact of the pandemic on its ability to accurately gather required information. Instead, the Bureau announced it would begin to issue “experimental” annual estimates, starting in November 2021. But it cautioned that those estimates should not be considered as replacements for the annual ACS totals.
In late July, the Census Bureau also issued a statement about the differences between the decennial Census totals and the ACS one-year estimates. The statement asserted the Bureau was able to “overcome pandemic-related challenges” in the decennial count, in contrast to the annual ACS survey, by “pouring additional resources into ensuring response.”
In recent days, however, the decennial count has come under increased scrutiny. On October 9, NPR reported that it had learned that the Census Bureau was planning to postpone and extend a follow-up quality-control survey for the decennial census, called the “Post-Enumeration Survey,” given growing concern about data quality. The Census Bureau’s “Press Releases” page on its website shows no announcement of this decision. And on October 13, in a story headlined “2020 Census may have undercounted Black Americans,” the Washington Post’s Tara Bahrampour reported that “the 2020 Census was fraught with challenges.” She included an assertion from a Democratic legislator from Michigan that “It was a perfect storm for an undercount on many levels,” and a statement from the CEO of the National Urban League that “this might be our greatest undercount since 1960, or 1950.”
Back in 2019, on the other hand, I took note of a developing “get-out-the-vote” campaign for the Census in Illinois. I questioned whether the Census Bureau had to manage the risk that some states, amidst intensifying fiscal stress, might be more motivated to “get out the vote” than others. More recently, on April 29, 2021, after the decennial results came in, I wrote an article titled “Did get-out-the-vote campaigns bias the Census -- and apportionment -- results?” I took a hard look at the differences between the “actual” decennial census totals and the annual ACS survey results. States with significantly higher 2020 decennial census results than those to be expected from the ACS estimates from 2010 to 2019 (before the pandemic hit) tend to be financially-challenged, and Democratic, states.
Meanwhile, a seemingly-reputable review by a task force established by the American Statistical Association recently concluded that it couldn’t reach a conclusion about the quality of the data delivered in the 2020 Census. It stated that “indicators released to date by the bureau do not permit a thorough assessment of the 2020 Census data quality.” Nonetheless, the task force report offered the following double-negative vote of semi-confidence: “Across the limited set of state level process statistics evaluated by the task force, it found no major anomalies that would indicate census numbers are not fit for uses of apportionment.”
Government accounting includes counting heads as well as dollars. And counting heads is not as easy as it may seem. Appraisals of the 2020 decennial census as well as annual population totals deserve close attention in coming months. There is a lot of money and power at stake in these numbers -- and where there is a lot of money and power at stake, political forces gather like bees on honey.