Symposium: There is no valid justification for the citizenship question

Wendy Weiser is the director and Kelly Percival is counsel of the Democracy Program at the Brennan Center for Justice at NYU School of Law.

At bottom, the citizenship-question cases rest on two principal charges: that the addition of a citizenship question to the 2020 census form sent to all households will lead to a significant undercount of the U.S. population, especially in immigrant communities and communities of color, and that Secretary of Commerce Wilbur Ross made the decision to add the question improperly, without adequately considering or testing its effects. According to the plaintiffs — and to the two lower courts that ruled in their favor — these charges compel the conclusion that Ross’ hurried decision to add a citizenship question was “arbitrary and capricious” in violation of the Administrative Procedure Act. They also claim it undermines the Census Bureau’s ability to fairly and accurately count the whole population in violation of the Constitution’s enumeration clause.

Rather than contesting those charges, the administration focuses most of its briefing in defense of the citizenship question on procedural issues of standing and reviewability. Substantively, it offers only two justifications for its decision to add the question: that it is merely reinstating a question that had previously long been used in the census, and that it needs the question to better enforce the Voting Rights Act. Neither justification has any merit.

History of citizenship questions on the census

The administration’s first defense of the citizenship question is a historical one. According to its briefs, every decennial census but one from 1820 to 1950 has included “questions about citizenship or country of birth (or both).” Far from a new innovation, the administration claims, Ross’ citizenship question “represents a return to the traditional status quo.” As a result, it argues, the question cannot be unconstitutional because such a ruling would “deem virtually every census questionnaire in the Nation’s history unconstitutional” — an improbable and intolerable result. Similarly, it argues, “it simply cannot be arbitrary and capricious” under the APA “to reinstate to the decennial census a question whose pedigree dates back nearly 200 years.”

The plaintiffs dispute the legal significance of this history. They argue that the citizenship question would dramatically undermine the accuracy of the 2020 census, regardless of what was done in the past; that there is no evidence that past citizenship questions did not similarly hurt past censuses; and that a past practice abandoned decades ago does not exempt Ross from conducting a proper analysis and following proper procedures before changing the census today. In other words, the plaintiffs claim, it is “arbitrary and capricious” to add a question to the census that hasn’t been used for over half a century without properly assessing its impact on today’s count. And any reasonable assessment would show that a citizenship question would cause a dramatic and disproportionate undercount of immigrant communities, in violation of Supreme Court precedent requiring census decisions to bear a “reasonable relationship to the accomplishment of an actual Enumeration” of the U.S. population.

These rebuttals should be sufficient to dispose of the administration’s 200-year-history argument as a matter of law. But the administration’s argument suffers from an even more fundamental flaw: Its historical account is wrong, or at least highly misleading.

First, it is simply not true that the census has previously asked for the citizenship status of everyone in the United States. According to new research published by our colleagues Brianna Cea and Thomas Wolf in the Georgetown Law Review Online and an amicus brief filed by leading census historians, there is no historical precedent — before 1960 or otherwise — for a universal citizenship question on the census. Rather, in past censuses, the government has either not asked about citizenship or has asked only a subset of the population. This matters because, as survey research shows, a question asked of some people has a very different impact on the count than a question asked of everyone.

Second, to the extent the administration relies on past instances of non-universal citizenship questions to justify Ross’ decision, its defense encounters another major historical problem: These questions were part of an approach to census-taking that the Census Bureau abandoned precisely because it determined that such an approach undermined the accuracy and efficiency of the count.

Before the mid-20th century, the census tried to do two things at once: enumerate the entire population and collect other information about U.S. residents (including at times the citizenship of some of those people). That approach, it turns out, did not work well. In the 1950s, the Census Bureau studied the accuracy of its count and found that its dual-purpose approach to the census, and the resulting longer questionnaire, caused millions of people to go uncounted, especially in minority communities. Accordingly, it changed its approach. Since 1960, it has used two separate questionnaires: one short, five-question form sent to everyone for enumerating the population, and one long-form questionnaire, sent to a smaller subset of the population to collect additional data.

The short-form questionnaire — the one at issue in these cases — has never included a citizenship question. And ever since the Census Bureau created a pared-down form, it has repeatedly strongly opposed any attempts to ascertain the citizenship status of everyone in the country because of the potential to jeopardize the accuracy of the decennial count.

In short, it is unreasonable to rely on census history to conclude that a citizenship question will not undermine the count.

The Voting Rights Act and citizenship data

The administration’s second justification — relevant only to the APA claims — is that collecting citizenship information on the decennial census is necessary for the Justice Department to enforce certain claims under the Voting Rights Act that require the use of citizenship data. The basis for this contention is a December 2017 letter from the Justice Department requesting the change. The administration argues that Ross’ decision was not “arbitrary and capricious” because he reasonably relied on this request and determined that “citizenship data provided to DOJ will be more accurate with the question than without it.”

The problem with this argument is that it too is inaccurate and implausible. As a threshold matter, the lower court found that Ross did not, in fact, rely on the DOJ letter — that he had made his decision well beforehand — and that his stated rationale of gathering better data to enforce the VRA was pretextual. But even if the Supreme Court refuses to look behind his stated rationale, it is not rational to assert that a citizenship question will improve VRA enforcement.

First, ever since the Supreme Court’s 1986 decision making citizenship data relevant to VRA claims, successive administrations and community advocates have successfully litigated VRA claims using citizenship data derived from other sources. Since 2005, the primary source has been the American Community Survey, a sample survey that replaced the long-form census and asks a subset of the population about their citizenship status. This data has been more than adequate for enforcing the VRA. Indeed, we are not aware of a single case in which the success of private plaintiffs in a VRA enforcement action turned on the availability of citizenship data from the decennial census. And the Justice Department has never, in the 54-year history of the VRA, cited a need for citizenship data from the decennial census.

Second, collecting citizenship information on the census would, as the lower court found “produce less accurate and less complete citizenship data” than other sources provide, thereby weakening — not strengthening — VRA enforcement efforts. What is more, according to the overwhelming consensus of census experts and the Census Bureau itself, the citizenship question will produce a significant undercount in minority communities and those with high immigrant populations. For example, the Census Bureau has estimated that a citizenship question would disproportionately depress response rates among Hispanic households. And bureau studies show that minority and immigrant households are less likely to respond to the 2020 census for fear that their responses might be used against them or their family members. The administration does not refute these estimates.

If the 2020 census reports artificially low minority population numbers as a result of the citizenship question, those communities will find it difficult to meet the preconditions required for making out a claim under the VRA using decennial census data. In other words, a citizenship question will make VRA enforcement more difficult for the very communities the statute is meant to protect.

Will it matter to the outcome of the case that there was no rational basis for Ross’ decision to add the citizenship question? It should. If the Supreme Court allows the administration to proceed with the question absent any plausible substantive justification, it will substantially weaken the APA and judicial oversight of agency decision-making. That, we hope, is a step too far for the justices.