Photograph by Nathaniel St. Clair
Back in February, the Trump administration announced sweeping overhauls that would radically weaken the Operational Test for the 2030 Census. The test is meant to serve as a dress rehearsal of sorts, allowing the Census to evaluate its operational mechanics before scaling them up nationwide. And it would appear that some of the administration’s most important changes are intended to undercount immigrant populations, though other populations will be affected as well.
As we noted in February, the administration sought several large-scale alterations that would undermine the utility of the test, including slashing the number of test sites from six to two and using Postal Service workers in place of trained Census Bureau enumerators to perform non-response follow-ups. The administration also opted to replace the standard short Census form with the much longer American Community Survey (ACS) questionnaire. The latter conveniently includes the same type of citizenship question that courts barred the administration from adding to the short-form. The Trump administration has further opted to move the ACS citizenship question earlier in the survey.
Despite substantial criticism, the Trump administration is moving forward with both the diminished suite of test sites and its plan to replace Census enumerators with mail carriers for non-response follow-up, a labor-intensive process of checking on households that do not respond on their own. The administration’s plan, which compromises the integrity of the Operational Test, presents a problem for overall Census validity that will fall hardest on groups that are already more likely to be undercounted.
First, some baselines. While the Census estimated that it ultimately enumerated 99.9 percent of all households in 2020, just under a third of households were counted through what the Census calls “non-response follow-up.” This process involves repeated in-person visits to households that do not respond on their own, along with careful, persistent fieldwork to determine who lives at a given address.
The remaining sites for the Operational Tests cover portions of Huntsville, AL, and Spartanburg, SC, though it is difficult to determine how, and to what extent, those included portions may differ from the parts of those metropolitan areas that are not part of the test. What we do know is that Huntsville had a 72.9 percent self-response rate, well above the national average of 67 percent. Spartanburg’s self-response rate was 63.7 percent. Though this is slightly below the national mean, it’s a far cry from Presidio County, TX, which had a self-response rate of just 27.4 percent in 2020. The Census had originally planned to include Presidio County and several of its neighbors in the test before the Trump administration moved to significantly restrict the number of test sites — thereby removing one of the few areas that would have meaningfully tested enumeration in hard-to-count communities.
The remaining test sites seem poorly suited to capture several of the populations most likely to be affected by these changes. The Census Bureau has repeatedlyidentified “migrants” in general and undocumented immigrants in particular as a hard-to-count population. Because the Census Bureau is testing only portions of Huntsville and Spartanburg, we cannot assume the sites themselves mirror metro-wide demographics. What we do know is that both Huntsville and Spartanburg have proportionally fewer immigrants than the US as a whole. Nationally, about 14.1 percent of residents are foreign-born, and 6.8 percent do not have US citizenship. By contrast, only 5.6 percent of Huntsville residents are foreign-born, and only 2.9 percent are not US citizens. In Spartanburg, those figures are 8.2 percent and 4.2 percent, respectively.
It is difficult to view the potential suppression of immigrant enumeration as accidental. The Trump administration has doggedly sought to exclude undocumented immigrants from the Census and apportionment process. The administration’s latest attempt to shoehorn a citizenship question is likely to depress responses from immigrant communities more broadly. The erratic treatment of refugees and Temporary Protected Status holders, combined with increasingly brutal ICE tactics, could give even many documented immigrants reason to fear sharing information with government officials. Under the circumstances, asking households to trust a Census process that is also experimenting with follow-up by postal workers — who are not held to the same confidentiality standards as Census enumerators — is likely to depress participation among a far broader swath of immigrants and mixed-status households. This is especially true if those workers are following up on a survey with a citizenship question, the mere prospect of which could depress participation among immigrant households and mixed-status families.
Substituting mail carriers for dedicated Census enumerators also threatens to exacerbate undercounts in rural areas and places with hard-to-reach or nonstandard addresses. Some households rely on PO boxes or cluster boxes rather than home delivery. Others are in remote areas, on tribal lands, or at addresses that do not map neatly onto standard postal routes. Several of the sites the Trump administration removed from the test — Western Texas, Tribal Lands Within Arizona, and Western North Carolina — were originally chosen to confront challenges in counting more remote populations.
A postal worker operating on an existing route also has far less time and flexibility than a dedicated enumerator to make repeated visits, ask probing follow-up questions, and work through confusion about who lives where. The households most likely to require persistent, specialized follow-up are the ones for whom mail-carrier replacement makes the least sense. Postal workers already provide a vital, time-sensitive public service, and they should not be asked to absorb an entirely different job that requires its own training, legal safeguards, and hours of additional work.
The administration is taking risks with the Census in ways that make those risks harder to detect before they are scaled up nationwide. The result could be a test that appears to succeed precisely because it does not adequately measure failure where it is most likely to occur.
This first appeared on CEPR.
