Census Smoke Signals

The decennial Census is a high-stakes feat. It informs government spending, business planning, and congressional apportionment. No wonder the Census Bureau takes such great pains with each question. After all, when you ask only ten questions or so, you've got to make them count (pun intended). You can almost hear the bureaucratic hand-wringing in the agency's requests for public comment: how should we word questions about race and Hispanic origin? Can we gather this information effectively through handheld computers, or through the internet? How will our "dress rehearsals" for collecting all this data go? For good reason, the Bureau engages in extensive survey pre-testing during the decade before show time. Census staffers must be an anxious lot.

One wonders, then, how Commerce Secretary Wilbur Ross--in a memorandum directing the Bureau to include a question on citizenship--could puzzle about whether a "mechanism" exists for learning about possible non-response rates. He is either seriously misinformed about how the Census Bureau usually works (troubling) or doesn't care (also troubling). Ross' claim that testing from the long-form American Community Survey, which already contains a citizenship question, is good enough also stretches credulity. You don't have to be an expert demographer to know that different surveys with different functions, contents, formats, lengths, consequences, and visibilities are going to produce different response rates.

The point of this post, however, isn't to wade into survey methodology, but instead to consider what is happening at the Census Bureau from the perspective of administrative law. Like any agency, the Bureau is subject to the Administrative Procedure Act (APA), as well as lesser-known statutes like the Paperwork Reduction Act and the Data Quality Act (more on these below). The APA requires that agencies refrain from acting in ways that are "arbitrary and capricious." Put differently, agencies must give credible reasons for their decisions grounded in expertise. Judge Leventhal of the D.C. Circuit has put the standard more colorfully: look out for "danger signals" that the agency has not really taken the problem seriously. In this view, courts should try to smoke out politicized intent unmoored from the available evidence (or lack thereof). Different judges can disagree on what these signals are, but in this case, they can take their pick:

Violations of an agency's own internal standards and procedures. Something is usually amiss when an agency doesn't follow its own procedures. That's the premise at least of the so-called Accardi doctrine. When an agency constrains its own discretion, it's legally obligated to operate accordingly. To be sure, the doctrine has the most bite when the agency has tied its own hands through a legislative rule, the violation of which results in third-party harm. But courts have often applied Accardi even in cases where the agency has issued an internal manual or other less formal means of commitment.

Here, the Census Bureau's own Methodology and Standards Council has promulgated statistical standards that explicitly apply to the decennial Census. Here is its statement of the most "minimal" standard for census questions:

The minimal standard is that a census or survey questionnaire must "work." This means that the questions (new or revised) can be administered properly by interviewers (if relevant) and understood and answered by respondents, and do not adversely affect survey cooperation. Evidence that a question works may be based on results of its use in prior surveys, or on tests with respondents conducted by the Census Bureau or by other organizations.

Evaluated accordingly, deciding to include a question about citizenship without engaging in any pre-testing seems to violate the Bureau's own standards for new questions. While this failure may not rise to the level of an actual Accardi violation (courts apply the doctrine in different ways and showing third-party harm will be tricky), the fact that the Bureau is not following its own publicly available guidance is highly suspicious.

Overruling career civil servants. Reports indicate that senior career officials at the Census Bureau opposed the addition of a citizenship question and "scrambled" to come up with alternatives. Apparently, this staff conclusion was informed by the agency's own observations that census responses could be affected by political messages that "certain immigrant groups are unwelcome."

Political appointees overrule career civil servants all the time. And under our system of democratically accountable agency control, it's good that they do. But civil servants are also often the keepers of agency expertise. In this case, they are the ones who have years of experience administering surveys. So when their judgments are cast aside, eyebrows should be raised.

Lack of senate-confirmed appointee. Often, when a political appointee of a parent agency like Commerce attempts to act in a highly politicized way, the head of a more expert subagency -- here that would be the Census Director -- will step in. Perhaps they will go toe-to-toe with that parent agency head in internal meetings or even go so far as to publicly protest (as happened with the FDA Commissioner objecting to a decision by President Obama's Secretary of Health and Human Services to prevent Plan B from being offered over-the-counter to certain age groups).

True, Congress has delegated the final decisionmaking power here to the Commerce Secretary, but the institutional reputation of Census is also on the line, and Census Directors usually care. Why isn't that internal check being exercised here? The Census has lacked a Senate-confirmed Director since last June. Acting officials can be great, but they lack the political capital and sense of legitimacy to challenge Senate-confirmed appointees.

Avoiding public comment. Finally, when an agency avoids its legal obligations to engage in public comment, the agency is either disinterested in the information it might receive or otherwise trying to ram a decision through. Here, at long last, we come to the Paperwork Reduction Act (PRA) and Data Quality Act (DQA). Suffice to say that both statutes require agencies to allow for public input on information integrity, with more formalized comment periods under the PRA. A quick search of the Federal Register, however, suggests that Commerce has not done anything of the sort for pre-testing of a citizenship question or related efforts to understand how such a question would impact the accuracy of the census. Unfortunately, neither the PRA nor DQA are judicially enforceable, but violations should be sounding political fire alarms.

Surveying the signals above, one can't help but think that the Commerce Department's decision was motivated by something other than the evidence before it. After all, the failure to collect evidence itself is usually a sign that some other motive besides expertise is at play. But this would not be the first time a judge has had to disentangle pretext from reality. And when a court is called upon to do so here, it would do well to remember that where there is smoke, there is usually a fire.