Document
260918- Generic Clearance for Census Bureau Internet-Based Testing
ICR 202610-0607-001 · OMB 0607-0978 · Object 174091000.
Document Viewer [pdf]
Status: Original and derived artifacts are available for this document.
Download: pdf
Loading document viewer…
Document Metadata
| File Type | application/pdf |
|---|---|
| File Title | 260918- Generic Clearance for Census Bureau Internet-Based Testing |
| Conversion State | complete |
Extracted Text
September 18, 2026 RE: July 20, 2026 Federal Register Notice on the Proposed Extension of Generic Clearance for Census Bureau Internet-Based Testing, Docket Number USBC-2026-0232 Dear Ms. Dumas: The Census Quality Reinforcement Task Force appreciates the opportunity to comment on the Census Bureau's request to extend its generic OMB clearance for medium-scale internet-based testing, including split sample experiments, remote usability testing, and qualitative research such as cognitive interviews, focus groups, and debriefings. CQR is a non-partisan, multi-disciplinary learning community for research, education, and coordination around decennial census data quality and fitness for use. Our members represent a broad range of expertise and experience, united by a shared commitment to census data quality. We support a robust internet-testing clearance in principle: rigorous pretesting of questionnaires, contact strategies, and data dissemination tools is essential to producing an accurate and cost-effective 2030 Census. However, our support is conditioned on the Bureau using this clearance to conduct testing that is methodologically sound, sufficiently transparent, and consistent with lessons the Bureau has already learned — and, in some cases, has recently disregarded. Our concerns are informed directly by the Bureau's handling of the 2026 Operational Test in Support of the 2030 Census, in which the Bureau eliminated critical field experimentation, narrowed test sites, substituted the American Community Survey (ACS) for the actual decennial questionnaire, and proposed English-only self-response testing. We are concerned that, absent clearer commitments, this generic clearance could be used to conduct additional testing activities that repeat those same shortcomings — but with even less public visibility, since activities under a generic clearance are reported to OMB in writing rather than subject to the same notice-and-comment scrutiny as a named operational test. These concerns have since been corroborated by an independent federal watchdog. In a June 4, 2026 report,1 the U.S. Government Accountability Office (GAO) found that the Bureau reduced 2026 Test locations from six to two and eliminated or reduced 10 of the 19 operational activities it had originally planned to test, including an internet self-response option for residents 1 U.S. Government Accountability Office. 2030 Census: Census Bureau Needs Additional Data to Inform Design Decisions. GAO-26-108848. Washington, DC: U.S. Government Accountability Office, June 4, 2026. https://www.gao.gov/products/gao-26-108848 of college dorms and other group quarters. GAO concluded that this narrowed scope creates a risk that the Bureau will finalize its 2030 Census design without adequate evidence, could lead to cost and quality challenges, and “may also harm congressional and public confidence in the Census.” GAO recommended that the Commerce Department instruct the Bureau to research and test the removed activities and design features before finalizing the 2030 Census design; Commerce agreed with the recommendation but, as of the report's publication, had not yet acted on it. We urge the Bureau to treat this generic clearance as one avenue for fulfilling that outstanding GAO recommendation. We offer the following recommendations, drawn substantially from concerns and recommendations we raised in response to the February 3, 2026 FRN for the 2026 Operational Test, to the extent they are relevant to internet-based testing conducted under this clearance. I. The Department's New English-Only Policy Does Not Override the Census Bureau's Statutory Multilingual Obligations We are aware that, effective June 4, 2026, the Department of Commerce issued Department Administrative Order (DAO) 201-46, “Designating English as the Official Language for Department Programs and Activities.” DAO 201-46 directs that the Department conduct “all domestic, non-public safety operations and activities in English—and only English—to the maximum extent permitted, consistent with applicable law.” Notably, DAO 201-46 also states that it “shall not be interpreted to conflict with any constitutional, statutory, regulatory, or other legal provision.” We raise this because it bears directly on the multilingual testing recommendations below. This is not a hypothetical concern. GAO's June 2026 report on the 2026 Census Test found that the Bureau had previously planned Spanish- and Chinese-language versions of the 2026 Test questionnaire but ultimately approved an English-only version. We are concerned that this generic clearance could be used to extend that same English-only approach to additional internet-based testing, notwithstanding the statutory language requirements described below. Congress has already spoken specifically to language use in Census Bureau data collection. providing that “the Secretary of Commerce shall ensure that, in the Bureau of the Census data-collection activities, the needs and concerns of the Spanish-origin population are given full recognition through the use of Spanish language questionnaires, bilingual enumerators, and other such methods as deemed appropriate by the Secretary.” This is a binding statutory directive specific to census data collection, not a matter of general agency discretion, and a general departmental administrative order – particularly one that expressly disclaims any conflict with existing statutory provisions – cannot narrow it. Multilingual census data collection also underlies the Bureau's ability to support determinations of the Voting Rights Act, which requires covered jurisdictions to provide election materials and assistance in the language of applicable Spanish, Asian American, American Indian, and Alaska Native language minority groups. The Census Bureau and the Department of Justice rely on Census Bureau data to make and update Section 203 coverage determinations. Testing and administering the census in English only would degrade the very data on which those separate statutory determinations depend. Finally, an undercount attributable to language barriers is a data-quality and accuracy problem, not merely a matter of inclusivity or accommodation. The Bureau operates under both a constitutional mandate to conduct an “actual Enumeration” and Title 13's broader mandate to pursue the most accurate count achievable. Multilingual testing and materials are a means of fulfilling that mandate, not a discretionary add-on that DAO 201-46's general English-preference policy displaces. II. Recommendations 1. Testing instruments should reflect the actual decennial questionnaire and platform, not substitute instruments. Split sample experiments and usability testing conducted under this clearance should use the actual decennial census questionnaire and Internet Self-Response (ISR) platform whenever the purpose of the test is to inform 2030 Census operations. Using a substitute instrument, such as the American Community Survey, introduces respondent burden and response patterns that differ meaningfully from the decennial context and will not reliably predict decennial self-response behavior. We recommend that any written notification to OMB describing a testing activity's purpose and scope specify the questionnaire instrument being used. 2. Multilingual testing should be standard practice, not an exception – and, for Spanish, is a legal requirement. Given that internet self-response testing may include tests of questionnaires, invitations, and websites, we recommend that this clearance not be used to conduct English-only testing of materials intended to inform decennial or demographic survey operations. As discussed in above, Spanish-language testing is required by law, independent of DAO 201-46's general English-only policy, and should not be scaled back on the basis of that policy. Consistent with 2020 Census practice, and to support accurate enumeration of language-minority populations more broadly, testing should be conducted in the full range of languages the Bureau intends to support in 2030 (at minimum English, Spanish, Chinese [Simplified], Vietnamese, Korean, Russian, Arabic, Tagalog, Polish, French, Haitian Creole, Portuguese, and Japanese). English-only testing under this clearance risks generating misleading self-response rate estimates, and in-field follow-up costs. We recommend that the Bureau publicly confirm – in its notifications to OMB under this clearance and in any internal guidance implementing DAO 201-46 – that Spanish-language testing and materials will continue as a matter of statutory compliance, and that testing in additional languages will continue to the extent necessary to support accurate enumeration, notwithstanding the Department's general English-only policy. 3. Split sample and usability testing should include populations and geographies affected by historical undercounts. The clearance notes that panel sizes and test parameters will depend on the objectives of each test. We recommend that, where a test is designed to evaluate contact methods, questionnaire design, or website usability intended for national deployment, the Bureau ensure test panels are not limited to convenience samples or a small number of metro areas, and that they include respondents from rural areas, tribal lands, and other areas with limited or no cell phone or internet service. Findings drawn from limited or non-representative panels should not be treated as generalizable evidence for 2030 Census design decisions. This concern is not speculative: GAO's June 2026 report found that the Bureau reduced 2026 Test locations from six to two, leaving Spartanburg, South Carolina and Huntsville, Alabama — two mid-sized Southern metro areas — as the only remaining sites, which undermines the representativeness of any results the Bureau, Congress, and the public will rely on to evaluate 2030 Census design decisions. 4. Contact-method testing should include the full range of strategies relevant to reducing undercounts. Testing of alternative contact methods, including emails and text messages, is one of the explicit purposes of this clearance. We recommend that such testing also examine variation in contact frequency, integration with mailed materials (e.g., QR codes), and targeted versus universal outreach strategies, since these variables materially affect self-response rates among historically undercounted populations, including children, rural residents, and low-income households. 5. Written notifications to OMB should include sufficient detail for meaningful oversight. The Bureau states that OMB will be informed in writing of the purpose and scope of each testing activity, along with the time frame and number of burden hours used. Given the recent narrowing of the 2026 Operational Test relative to its originally announced scope, we recommend that these written notifications include, at minimum: the questionnaire or instrument used, the languages tested, the composition and size of the test panel, the geographic scope of the panel, and whether the activity is intended to inform decennial census operations specifically. This level of detail would allow for meaningful post hoc review of how the clearance is being used. This recommendation is reinforced by GAO's own experience: in preparing its June 2026 report, GAO sought a rationale from the Department of Commerce for the changes made to the 2026 Test's scope, but as of the report's publication, Commerce had not responded. We recommend that written notifications under this clearance include an explanation of any material changes to previously planned test scope, panel composition, or geographic coverage. 6. Usability and qualitative testing should include accessibility and limited-English-proficiency populations. Remote usability testing and qualitative interviews conducted under this clearance should be designed to capture the experiences of respondents with accessibility needs (e.g., screen reader users, individuals with hearing impairments) and respondents with limited English proficiency, including “passive bilingual” individuals who understand English but cannot easily complete a questionnaire in English. Excluding these populations from usability and cognitive testing risks propagating design flaws that disproportionately affect groups already at higher risk of being undercounted. 7. Quality control measures should extend to internet-based testing activities. We recommend that split sample experiments and usability tests conducted under this clearance include basic data quality safeguards, such as attention checks, response-time analysis, or validation samples, to ensure that results are reliable enough to inform decennial design decisions. The absence of quality assurance procedures in a test undermines confidence in any resulting operational recommendations. 8. Test results should be published publicly, not reported only to OMB — and internal memos are an acceptable substitute for polished reports where capacity is limited. Public release of testing results, not just internal reporting to OMB, is essential to enable outside review of the Bureau's testing program, particularly given the reduced visibility of activities conducted under a generic clearance, described above. The Bureau has a documented pattern of not publishing results even when testing is completed in recent years. Several studies from the 2020 Census Evaluations and Experiments (EAE) program were conducted but never released: the Bureau's own EAE program page states, relative to 9 completed evaluations and experiments, “the Census Bureau conducted this evaluation, [but] it will not externally release the accompanying report.”2 A related February 2024 Inspector General report further found that a number of 2020 operational assessments, evaluations, and experiments were not completed in time to formally inform the 2030 Census research and testing agenda. Similarly, results for tests listed on the Bureau's 2030 Census research and testing pages have gone unpublished well past their expected release; the 2023 Census Test webpage, for example, has stated since October 2023 that a results report “will be” posted “once available,” but no such report currently appears there. We recommend that the Bureau commit to publicly releasing results for every testing activity conducted under this clearance, on a rolling basis as each test concludes, rather than relying solely on written notifications to OMB. Where the Bureau faces capacity constraints that make producing a polished, external-facing report infeasible for a given activity, we recommend that the Bureau instead release the internal memorandum or analysis it already prepared for its own or its sponsor's use. Releasing an existing internal work product is far less resource-intensive than drafting a new public report, and it would spare researchers, advocacy organizations, and other interested parties from having to file Freedom of Information Act requests to obtain results the Bureau has already produced for its own purposes. In conclusion, we recognize that a generic clearance mechanism offers the Census Bureau necessary flexibility to conduct ongoing, iterative testing over a multi-year period. That flexibility, however, should not come at the expense of methodological rigor or public accountability — particularly given the Bureau's recent decision to narrow the scope of 2026 Operational Test field experimentation in ways that will undermine the accuracy of the 2030 Census. We urge the Bureau to incorporate the recommendations above into its use of this clearance, and to ensure 2 www.census.gov/programs-surveys/decennial-census/decade/2020/planning-management/evaluate/eae.html that internet-based testing conducted under it produces evidence that is genuinely representative of the populations and conditions the 2030 Census must accurately enumerate. We appreciate the opportunity to comment and welcome further engagement with the Bureau on these issues. Sincerely, Allison Plyer Co-Chair of the CQR Task Force Chief Demographer, The Data Center Dr. Jeanine Abrams McLean Co-Chair of the CQR Task Force President, Fair Count Cara Brumfield Co-Chair of the CQR Task Force Senior Fellow, National Conference on Citizenship