Robert L. Santos is vice president and chief methodologist at the Urban Institute and president-elect of the American Statistical Association.

The Supreme Court heard oral arguments Monday on the Trump administration’s bid to exclude undocumented immigrants from population totals when calculating how House seats are apportioned among the states. This is the latest threat to the quality of the 2020 Census that should worry both the public and the Census Bureau.

The 2020 Census has faced unprecedented challenges, including funding shortfalls, natural disasters, high employee attrition and litigation. Census field operations were compressed first by pandemic lockdowns and shortened further by court order. Millions of Americans relocated because of covid-19, intensifying challenges to an accurate count. Typically, processing census data takes about five months. In this chaotic year, data is supposed to be reconciled — with duplicate responses and other errors identified and corrected — over just 10 weeks. Shortcuts are unavoidable if counts are to be submitted as currently required by Dec. 31, elevating the risk of errors.

No one should want that. Census inaccuracies risk constitutional conflict, with decade-long ramifications for political representation and federal funding. Given the stakes, the Census Bureau should be relieved of the pernicious burden of submitting counts by year’s end.

So far, indicators of 2020 Census performance provide a mixed picture. The Census Bureau has touted a 99.98 percent “resolution” rate of all U.S. housing units and addresses as evidence that the forthcoming count is complete. But this figure relies on data of varying quality. It draws on federal and commercial administrative records, such as tax documents, that could be obsolete; it includes data from proxies such as neighbors or landlords to fill in gaps where census enumerators couldn’t reach households.

The self-response rate for the 2020 Census — households that completed their own responses, whether online, by postal mail or telephone — is 67 percent. That exceeds the 60.5 percent rate the bureau had hoped for, a notable achievement during a pandemic. Self-responses provide the most accurate information. Lower self-response rates are associated with lower-quality reporting and net undercounts.

But the self-responses, as illustrated in the “hard to count” map, also reveal a complex story. Across metro areas, suburbs exhibit self-response rates of 70 to 80 percent, with some ranging even higher. Meanwhile, core urban neighborhoods — mostly in communities of color — often have self-response rates below 50 percent. These disparities are also evident in hard-to-reach rural counties across the Deep South and the West, regardless of their populations’ racial and ethnic composition. Historically, lower self-response rates among minority households produced net undercounts of those groups.

In other words, despite the rosy picture suggested by the overall rates, there is a clear risk that people of color in urban areas and some rural communities will again be undercounted, with decade-long implications for fair political representation and federal funding. The challenges to this year’s census, discrepancies already observed in community self-response rates and the condensed processing period combine to heighten the risk of undercounts.

The Trump administration’s bid to exclude undocumented immigrants from calculations of congressional representation is another threat to census integrity. It risks conflict with the Constitution’s direction of “counting the whole number of persons in each State” and could exacerbate inequities in federal funding and political representation.

Warning lights have been flashing for a while. After recent reports that the Census Bureau had identified data issues that would delay delivery of a final count past the Dec. 31 deadline, the bureau’s director, Steven Dillingham, confirmed the discovery of anomalies in a one-paragraph statement. The nature and extent of these problems are not public, and the Commerce Department declined to respond to congressional requests for details.

More transparency is critical, not just for researchers but also to ensure public trust in the results. The American Statistical Association and the Census Bureau’s Scientific Advisory Committee have submitted independent reports urging transparency and recommending the release of quality indicators to help clarify the strengths and blemishes of final counts.

Dillingham got it right when he said that “our goal remains an accurate and statistically sound Census.” He has directed “all resources available” to address the discovered anomalies “as expeditiously as possible,” but the deadline has not been formally changed.

Census Bureau staff have reportedly recommended a Jan. 26 submission date so they have time to complete their due diligence. This means the Biden administration would oversee final steps, but there should be nothing political about this process. It is imperative that census staff be given sufficient time to process the data to acceptable levels of quality. Like all federal workers, Census Bureau staff swear an oath to support and defend the Constitution. This pledge and their commitment to scientific integrity are crucial to our democracy. If census staff are not given more time, then they need to be true to their constitutional oath and their scientific obligation to ensure the integrity of the counts — even if that means missing their deadline.

Census Bureau career staff are best-positioned to know whether the counts will be trustworthy. They shouldn’t compromise their standards — or compromise the 2020 Census.

Read more: