Census undercounts are historic — and bipartisan

The career scientists and professionals at the U.S. Census Bureau no doubt cringed at the thesis expressed in the June 1 opinion piece by Kristin Tate in The Hill and its most unfortunate headline. It is a shame that the bureau’s consistent effort to provide a report card after every decennial, not just 2020, is used in this case to advance a partisan political theory regarding the 2020 outcome, especially following a historic pandemic that disrupted census operations and public cooperation with the effort.

The nation’s “largest peacetime mobilization” is always more than 10 years in the making, incredibly complex, and sensitive to all kinds of disruptive factors, including recessions, natural disasters, variable civic promotional efforts across states, and yes, even a pandemic. The fact is, it is not possible for census managers to introduce a partisan sleight of hand on such a massive undertaking, even if they were not the dedicated scientists, interested only in ground truth, that we know them to be.

No American census is perfect. Trust us, we know, having held leadership positions across eight presidential administrations — four Republican and four Democratic — since the 1980 count. The Census Bureau knows this better than anyone. Bureau employees issue a post-census report card — formally known as the Post Enumeration Survey, or PES — because they are dedicated to making the next census better, more accurate and more complete than its predecessor. 

For the most part, they have succeeded across 24 national head counts since the first, in 1790. However, because of a number of factors — including the COVID-19 pandemic and unprecedented efforts by the Trump administration to add a citizenship question to the 2020 census questionnaire — that trend was broken in 2020. It disappoints the nation no less than the census professionals, but in fairness, not as much as most feared given the difficult challenges they faced. 

It’s true that 14 states were miscounted in 2020 by statistically significant margins. Overall, the PES reported a net national undercount in 2020 of just 0.24 percent. By comparison, the 2010 census, often regarded as the best count in the history of the process, had a net national overcount of just 0.01 percent, or almost zero. The 2000 census had a net national overcount of 0.49 percent. These three outcomes are in a very close range, especially when compared to the 1990 census, which had a national undercount of 1.61 percent, and every state in 1990 had an undercount of varying sizes.

The large undercounts measured for the 1990 census outcome motivated the bureau, based on research, to ask Congress to fund new tools, especially the first paid advertising and promotion effort to build upon the public service ads of the 1980 effort to motivate higher levels of civic engagement and cooperation with the 2000 count. It worked. The bureau reversed a three-decade decline in census form response rates.

George Washington and Thomas Jefferson, who was Washington’s secretary of State, reportedly expressed skepticism with the 1790 count, questioning whether it seriously undercounted the size of the fledgling nation. But in that instance, and in every census since, the nation accepted the results. They were fit for use, as are the 2020 results.

The nation survived simply fine after the disappointing 1990 results with its large undercount in big cities and among minorities. No effort to redo the apportionment of Congress for the following decade was undertaken then, and none is needed now.

Vince Barabba is the only Census Bureau director appointed by presidents of both parties, serving from 1973-1976 and 1979-1981, when he oversaw the 1980 decennial count. He also served as commissioner of the California Citizens Redistricting Commission.

John H. Thompson was the 24th director of the Census Bureau, from 2013-2017, leading the planning for the 2020 decennial. He also served in various senior leadership positions at the bureau overseeing work on the 1990 and 2000 counts.

Editor’s note: This article was edited after publication to clarify the 2020 and 2000 percentages of net undercount/overcount.