XXI. Minimum Guardrails for a Fair and Accurate 2030 Census
The 2030 Census design will continue to change. New research will be completed, tests will produce results, operational rules will become more specific, and some decisions may be revisited in response to policy, funding, litigation, or technical constraints.
Advocates therefore need more than positions on individual proposals. They need a set of durable standards against which new decisions can be evaluated.
The guardrails below are intended to serve that purpose. They do not prescribe every operational choice Census should make. They identify the minimum conditions that should hold across those choices if the 2030 Census is to remain accurate, equitable, transparent, and worthy of public confidence.
A. Direct Participation Must Remain the Preferred Method
Census should continue to design the operation around obtaining information directly from households whenever reasonably possible. Administrative records and other outside data can be valuable supplements and backstops, but they should not become a shortcut for avoiding meaningful contact.
Advocates should press for a clear presumption in favor of direct response, defined minimum contact expectations, and public reporting on when and why Census departs from that presumption.
B. Historically Undercounted Communities Must Receive More, Not Less, Field and Outreach Support
Efficiency tools should be used to concentrate resources where they are most needed, not to withdraw effort from households that models predict will be difficult, expensive, or unlikely to respond.
Advocates should insist that Census evaluate new models and operational strategies by asking whether they increase effective contact and response among historically undercounted populations. Savings achieved by reducing effort in the communities already at greatest risk of omission should not be treated as successful modernization.
C. Multilingual Access Must Be Preserved Across Modes
Language access should remain a core enumeration strategy rather than an optional communications service. People should be able to understand the census and respond through meaningful multilingual options across the principal response modes.
The advocacy objective should be straightforward: preserve robust multilingual questionnaires, internet and telephone response, field support, mail materials, and partner resources. Any proposed reduction should be justified with evidence about its effect on response and data quality, not simply by reference to a department-wide language policy.
D. Core Demographic Content Must Remain Scientifically Valid and Useful
Questionnaire content should be governed by statistical standards, research, testing, legal requirements, and sustained public engagement. Political disagreement with the existence or visibility of a population is not a valid basis for degrading the data.
Advocates should defend a process as much as a particular questionnaire. Changes to race, ethnicity, household relationships, or other consequential content should require a documented evidentiary basis and meaningful opportunity for scientific and public review.
E. Self-Response Should Receive Presumptive Priority
When Census receives information directly from a household, that response should ordinarily control over conflicting administrative or third-party information.
Departures may sometimes be necessary to resolve duplicates, residence questions, obvious errors, or conflicting submissions. But those exceptions should be narrow, rule-based, documented, and auditable. Advocates should seek a clear hierarchy of evidence rather than allowing opaque algorithms to determine silently when a person’s own response is displaced.
F. Administrative Data Must Be Evaluated for Differential Quality
A dataset should not be considered “high quality” merely because it contains many names and addresses.
Census should evaluate every important source for the particular purpose for which it will be used and for the populations most likely to be poorly represented in it. That includes residence, household composition, race and ethnicity, relationships, and other characteristics, not simply record coverage.
Advocates should push for published differential-quality assessments and should oppose national averages as sufficient evidence when a source performs unevenly across communities.
G. In-Office Enumeration Must Be Constrained and Transparent
If In-Office Enumeration (IOE) becomes a meaningful source of the 2030 count, it must be possible to evaluate where, how, and for whom it was used.
Census should commit in advance to publishing the share of people and households counted through IOE, the contact attempts that preceded it, the types of records used, and the resulting quality measures. Reporting should identify important geographic and demographic patterns rather than providing only a national total.
Advocates should also seek firm limits on when IOE is permissible. Its legitimacy depends on remaining a carefully validated backstop, not becoming an invisible alternative census for populations that are harder to reach.
H. Coverage Measurement Must Remain Genuinely Independent
The system used to evaluate the census must be capable of detecting the census’s mistakes.
As Coverage Estimation incorporates more administrative data and automated matching, advocates should focus on independence of evidence, not simply organizational separation. Census should demonstrate that the measurement system can identify people missed because administrative records were weak, addresses were absent, matching failed, or operational assumptions were wrong.
Where the census and the coverage system share data sources or methods, Census should explicitly measure the risk of shared error and preserve independent field collection sufficient to test those assumptions.
I. Disclosure Avoidance Must Preserve Usable Local Data
Confidentiality and data usefulness should both be treated as requirements, not as competing goals in which one can simply eliminate the other.
Advocates should avoid becoming locked into defense of any single technical method. The stronger position is to demand measurable standards for both privacy protection and data utility and to require Census to demonstrate that the chosen system preserves the local geographic and demographic detail necessary for redistricting, civil rights, public administration, and community use.
Demonstration products must come early enough for those claims to be tested independently.
J. The Public Must See Consequential Rules Before Implementation
Transparency is most valuable while decisions remain changeable.
Census should disclose major operational specifications, thresholds, model uses, data-source decisions, and research findings when they become sufficiently developed to drive requirements, procurements, or software. Advocates should not accept a process in which the public sees only high-level commitments until the underlying rules have already been built into production systems.
The practical standard should be simple: if a decision is consequential enough to shape how people are counted, it should become visible before changing it is operationally impractical.
Using These Guardrails
These standards can provide a common agenda even when advocates disagree about individual technical choices. They also create a way to evaluate decisions that have not yet been proposed.
For each major 2030 development, advocates can ask:
Does it preserve a meaningful opportunity for people to participate directly?
Does it improve or weaken the count for populations historically at greatest risk?
Is there evidence that the method works equally well across communities?
Can the public understand and evaluate the rule before it becomes fixed?
Will Census publish enough information afterward to determine whether it worked?
Not every dispute will have an obvious answer. But these questions shift the debate away from whether a particular innovation is inherently “modern,” “efficient,” or “data-driven” and toward the outcomes that matter.
The goal for the remainder of the decade should not be to preserve the 2020 Census unchanged. It should be to ensure that modernization strengthens rather than weakens the basic commitments underlying a fair census: everyone has a meaningful opportunity to be counted, additional effort reaches the people most likely to be missed, information is collected and used according to sound statistical principles, and consequential decisions remain open to public scrutiny before they become irreversible.