With all of the partisan fights over absurdities like the President’s dalliances with a porn star, Facebook personality quiz apps, and whether the President can find a competent lawyer, one would think that a subject as anodyne as the decennial census would be safe from inter-party bickering. Unfortunately, that’s no longer the case, as seemingly every aspect of government is being fought over these days, even the incredibly boring (although important) ones.
The new controversy over the census has been brewing for some time, but truly exploded on Monday when Commerce Secretary (and Mr. Magoo look-alike) Wilbur Ross detailed the department’s plan to add a question on citizenship status to the census documentation. Since the 1960 version of the census, no questions on citizenship status have been included on the regular census document, which goes out to all American households every ten years. Prior to that time, citizenship status was asked about in various ways for at least the preceding 100 years on the census itself. In 2000, the long-form census questionnaire, which went out to 1 in every 6 people in the US (everyone else got the shorter version), included a question on citizenship. The American Community Survey (ACS), the replacement for the long-form census adopted after 2000, includes questions on citizenship status and is sent to approximately 2.6% of the population on an annual basis. All of these prior and current census documents included citizenship information, so the precedent is there for asking these sorts of questions. (As a sidebar, the United Nations promotes including citizenship questions on censuses around the world.)
Why would someone want to include a question on citizenship status in the first place? Well, ascertaining the citizenship status of those living in the United States is an important policy goal, as learning how many immigrants (legal and illegal) live in the country, and where, is useful in better allocating funds and supporting those populations through provision of government services. The justification for the addition of the new question to the 2020 census, however, is not focused on that rationale, instead building a case based on the Justice Department (DoJ) stating that they need census-block-level voter data to ensure enforcement of the Voting Rights Act (VRA) sections focused on promoting minority voting rights. As the Justice Department has been fairly lax in that sort of VRA enforcement thus far, and has taken the side of those supporting gerrymanders in the Supreme Court, I am doubtful that this motivation is legitimate. Many other commentators agree. Not only writers and pundits are concerned, as multiple states have already filed suit against the Trump administration for adding this question.

What are these commentators and states worried about? They claim that inclusion of a census question on citizenship status will result in large percentages of undocumented immigrants as well as legal immigrants simply avoiding or not completing the census; this would create a systematic undercounting of residents of areas with high immigrant populations (generally urban areas that predominately vote Democratic). It is worth delving into this argument in greater detail, as the evidence for this is mixed and inconclusive. According to the memorandum produced by Secretary Ross, response rates for the ACS (which includes a citizenship question) fell 3.1% below the response rates of the 2010 decennial census, but this difference could be explained by the lack of follow-up for the ACS compared to the full census, as well as the longer nature of the survey. The memo goes on to discuss response rates to the specific citizenship question on the ACS; on average between 6% and 12.6% of respondents did not complete that particular question on the 2013-2016 surveys. The Secretary also states that Nielsen data shows no material difference when they asked the same citizenship questions that would be on the census, compared to when they chose not to ask those questions. (I don’t find the Nielsen example compelling, as the responses to a private survey company and the American government are not directly comparable, but chose to include it as Ross uses it to bolster his argument.)
In the Ross memo, he discusses multiple options to responding to the DoJ’s request for data to help enforce the VRA; one option is to simply not respond or change the census (Option A in the memo), another is to add the citizenship question (Option B), and the third is to rely on US government administrative data to provide the information the DoJ requires (Option C). It is clear to me that Option C would be the best choice here as it allows the DoJ to get the data it needs (from a more reliable source, as administrative data is taken from other governmental agencies instead of relying on self-response surveys like the census) and avoids putting a question on the census that could reduce response rates. Ross chose none of these options, instead creating an Option D, which was a mix of Options B & C, both adding the question to the census and working to achieve the results the DoJ requested through administrative data. His rationale was that using administrative data alone would require the Census Bureau to impute the citizenship status of “more than 10% of the American population,” which would be too inaccurate. This is undercut by the fact that I (and Secretary Ross) mentioned earlier, which shows that the inclusion of a citizenship question reduces response rates on the ACS by between 6% and 12%. He also ignores the fact that “between 28 and 34 percent of the citizenship self-responses for persons that administrative records show are non-citizens were inaccurate.” (Quote from Ross’s own memo.) These mistaken responses, combined with the prior experience with ACS non-responses, show that relying on census data would still allow errors to exceed the 10% threshold Ross claims is too high.

Now that we’ve gone through all of the statistics and complex arguments around how citizenship data should be ascertained and used for Justice Department purposes, we need to get to the roots of the issue: what is the census supposed to be for and why are accurate counts even important in the first place? The basic purpose, and process, of the decennial census is laid out in the 14th Amendment to the US Constitution, which states (emphasis added):
Representatives shall be apportioned among the several states according to their respective numbers, counting the whole number of persons in each state, excluding Indians not taxed.
This clearly shows that all persons, immigrant or not, legal status or not, need to be counted for the purpose of the decennial census. Potentially reducing response rates by adding citizenship questions would belie the original purpose of the census altogether, as it would not render an accurate count of “the whole number of persons in each state.” Not only does an inaccurate count go against the census’s Constitutional purpose, but it severely negatively impacts all those academics, researchers, and businesses that rely on census data to make important decisions and track trends. Census data drives not only Congressional apportionment (revision of number of seats in Congress for each state), but provides the data which so many people and corporations use to learn about their fellow Americans. Census data is incredibly important for these groups to properly target their resources and make economic decisions. Inaccurate or incomplete data would seriously hamper the critical decision-making process. (I find this non-partisan argument far more persuasive than the argument many other opponents of the added question make, which is based on the administration trying to screw over Democrats in Congress. Generally, non-partisan arguments should take the fore, as they have the potential to convince many more people.)
The decennial census is a long-standing and crucial service that the government has carried out since our founding (the first census was all the way back in 1790). There were already significant issues surrounding this upcoming census, including lack of appointees to run the process, budgetary shortfalls, and poor implementation of intergovernmental recommendations to improve the process; we don’t need anything else potentially impacting the accuracy of data gathered through this incredibly vital survey.