Sergio M. Marxuach

PDF Version

U.S. Census Bureau
U.S. Department of Commerce
4600 Silver Hill Road
Designation: ADDEMO/FRN
Washington, DC 20746

Submitted electronically via the Federal eRulemaking Portal (www.regulations.gov)

Re: Comments of the Center for a New Economy on the Proposed Rule, “Decennial Census of the Population of Americans; Proposed Residence Criteria and Proposed Regulations for Demographic Questions,” 91 Fed. Reg. 57,524 (Sept. 10, 2026); Docket No. USBC-2026-0628; RIN 0607-AA75

To the Census Bureau:

The Center for a New Economy (“CNE”) respectfully submits these comments on the proposed rule that would add 15 C.F.R. Parts 60 and 70 to govern the decennial census (the “Proposed Rule”).^[1] CNE is a not-for-profit, nonpartisan think tank based in San Juan, Puerto Rico. Our research on Puerto Rico’s public finances, health care financing, energy, and federal–territorial affairs relies on decennial census counts and on the statistical products built on them, including the Puerto Rico Community Survey (“PRCS”) and the Census Bureau’s annual population estimates.

CNE opposes the Proposed Rule as drafted. The more than three million residents of Puerto Rico are enumerated in every decennial census under the same statutory authority the Proposed Rule invokes, yet the Proposed Rule does not mention Puerto Rico once, and the accompanying Preliminary Regulatory Impact Analysis (“PRIA”) mentions it only in passing.^[2] That omission is not a mere technicality. It leaves the public unable to determine which rules will govern the 2030 Census of Puerto Rico, and the Bureau has given no indication that it considered how its two central innovations—(1) a day-count residence standard “evidenced by” tax records, which the preamble indicates the Bureau would obtain from the IRS, and (2) a prohibition on race and ethnicity questions—would operate in a jurisdiction where most residents owe no federal income tax on Puerto Rico–source income and many file no federal return, where a large share of addresses cannot be reliably matched to a dwelling, where tens of thousands of people move between the island and the States each year and many more move back and forth within the year, and where the decennial census is the only full-count source of demographic data. The count that results will set, for the next decade, the population base for Puerto Rico’s legislative districts, the PRCS, federal formula funding, and the island’s certified fiscal plan.

These comments address three specific defects of the Proposed Rule:

1. The Proposed Rule does not state whether or how Parts 60 and 70 apply to Puerto Rico.
2. The residence standard in proposed § 60.2, a day count “evidenced by” tax records, is structurally mismatched to Puerto Rico’s tax system, address frame, and migration patterns.
3. Proposed Part 70 would eliminate Puerto Rico’s only full-count race and ethnicity benchmark, and could thin its small-area tables, without any analysis of the impact on Puerto Rico.

These comments are limited to the Proposed Rule’s effects on Puerto Rico. CNE’s silence on other aspects of the Proposed Rule should not be construed or interpreted as agreement with them.

1. The Proposed Rule Does Not State Whether or How It Governs the Census of Puerto Rico

Congress has directed that each census authorized under chapter 5 of Title 13 of the U.S. Code shall include the Commonwealth of Puerto Rico.^[3] Puerto Rico has been included in every decennial census since 1910.^[4] The Proposed Rule rests on 13 U.S.C. § 141, which is part of that same chapter.^[5]

The Proposed Rule’s justification, however, is framed almost entirely around apportionment among the States. The preamble describes apportionment as the “sole Constitutional purpose” of the decennial census;^[6] proposed § 60.1 defines the residence criteria as those needed to count each individual “for purposes of apportionment”;^[7] and proposed § 70.1 justifies the question restrictions by reference to “congressional apportionment.”^[8] None of these rationales reaches Puerto Rico, which has no apportioned seats.

The operative text, by contrast, is broader than its rationale. The summary describes residence criteria that determine where to count people “throughout the United States,” and proposed § 70.2(a) extends the question ban to “any questionnaire used for the enumeration of the population.”^[9] One purpose the rule itself names also cuts against the apportionment framing. Proposed § 70.3(a) directs the Bureau to prioritize the accuracy of the count “for purposes of apportionment and (re)districting,”^[10] and Puerto Rico’s Constitution requires that its legislative districts be revised after each decennial census.^[11] The rule thus recognizes at least one use of the count that applies fully in Puerto Rico, yet says nothing about whether or how it would apply to Puerto Rico.

The Proposed Rule also abandons the definitional clarity of the criteria it would replace. The 2020 residence criteria defined “stateside” as the 50 States and the District of Columbia, expressly classified Puerto Rico as “outside the United States” solely for purposes of those criteria, and noted that redistricting data include Puerto Rico’s resident population.^[12] The Proposed Rule retains the term “stateside census” in proposed § 60.5(a), but defines neither “stateside” nor “United States” and offers no explanation for dropping the prior definitions.^[13]

At least three readings of the Proposed Rule are therefore available: (i) Parts 60 and 70 govern the Puerto Rico enumeration in full; (ii) neither part applies, and the enumeration will be governed by criteria that have not been published; or (iii) Part 70 applies through its “any questionnaire” language while Part 60, limited by its apportionment purpose, does not. The preamble’s regulatory analyses do not resolve the question; they address small governmental entities, States, and tribes, but not territories.^[14] Under any of these readings, the Proposed Rule would affect how Puerto Ricans are counted: at a minimum, Part 60’s day count would decide whether residents who divide their time between the island and the States are counted in a State or in Puerto Rico (Part II.C).

The only indications of the Bureau’s intent appear outside the rule, in the PRIA, which observes that “the decennial census includes all 50 states, Puerto Rico, the District of Columbia, and the U.S. Island Areas” and states in a footnote that the Bureau conducts the census in Puerto Rico “operationally the same as the stateside decennial census.”^[15] The same footnote describes Puerto Rico as “an Island Area under 13 U.S.C. § 191,” a characterization the statute does not support. Section 191 does not use the term, and it treats Puerto Rico differently from the Virgin Islands, Guam, and the Northern Mariana Islands by omitting it from subsection (b), which allows the Secretary to rely on census information collected by those territories’ governors.^[16] The PRIA’s very next footnote lists the Island Areas without listing Puerto Rico.^[17] If the Bureau intends Parts 60 and 70 to govern the Puerto Rico enumeration, then the rule itself must state so clearly.

CNE therefore respectfully requests that the Bureau:

1. State expressly whether, and to what extent, Parts 60 and 70 govern the 2030 Census of Puerto Rico.
2. If the Bureau intends either part to apply, publish a supplemental notice that addresses the Puerto Rico–specific issues set out in Parts II and III of these comments and invite comment on them.
3. If neither part is intended to apply, publish for notice and comment the residence criteria and questionnaire content that will govern the Puerto Rico enumeration, with definitions of “stateside” and “United States” at least as clear as those in the 2020 criteria.

2. The Residence Standard in Proposed § 60.2 Is Structurally Mismatched to Puerto Rico

To the extent Part 60 governs the census of Puerto Rico, as the PRIA’s description of the Puerto Rico enumeration suggests, its proposed residence standard would fail there for the reasons that follow. Proposed § 60.2(a) provides that a person’s usual residence “should be consistent with, and evidenced by, their tax records (e.g., tax returns, W-2 forms).” The Bureau states that the weight it assigns to tax and other administrative records will be “situation dependent,” taking into account each data element’s “purpose, timing, and reliability.”^[18] Puerto Rico is a situation in which that standard has unusually little to work with, for two independent reasons, and the day-count test to which that standard is attached raises a third problem, discussed in Part II.C.

A. Residents’ Puerto Rico–source income is, by design, outside the federal income tax, and many residents file no federal return.

Under section 933 of the Internal Revenue Code, income from Puerto Rico sources is excluded from the federal gross income of bona fide residents of Puerto Rico.^[19] Such residents generally must file a federal income tax return only if their income subject to U.S. tax (income from sources outside Puerto Rico or wages as a federal government employee working in Puerto Rico) exceeds the applicable filing threshold, or if they wish to claim certain federal tax credits. Residents report their Puerto Rico–source income instead to the Puerto Rico Department of the Treasury (“Hacienda”).^[20]

Some island residents do file federal returns, but federal filing turns on income source and credit eligibility, not on where a person lives. For a resident whose income is entirely from Puerto Rico sources, the absence of a federal return is the expected result and says nothing about their residence. Since tax year 2021, moreover, families with children have had reason to file to claim the refundable Child Tax Credit,^[21] so federal returns now cover some households, such as families with children, far better than others, such as retirees and workers without children. That makes them a biased measure of residence.

The Proposed Rule does not address this issue. The only legal channel for tax data it describes is 26 U.S.C. § 6103(j)(1), which governs federal returns and return information.^[22] Its list of the records available to the Bureau includes “certain state, local, and tribal records,” but is silent as to territorial records.^[23] Although Title 13 permits the Secretary to acquire records from other units of government,^[24] the Proposed Rule does not say whether the Bureau holds or intends to seek Hacienda data, under what agreement, or with what coverage.

The principal wage record that crosses the federal–territorial line is Form 499R-2/W-2PR, which employers in Puerto Rico file with Hacienda and transmit to the Social Security Administration.^[25] The Proposed Rule’s reference to “W-2 forms” does not state whether Form 499R-2/W-2PR qualifies. Even if it does, it covers only wage employment. It says nothing about residents without wage income during the enumeration period, such as retirees and dependents, or about workers in the informal economy. One econometric study estimated that sector at about 36 percent of Puerto Rico’s GDP in 2014, and a recent field survey of low-income communities found more employment than official statistics report.^[26]

Notably, the Bureau’s own assessment of the 2010 Census of Puerto Rico recommended that the Bureau “[r]esearch the coverage of administrative records” for the Puerto Rico enumeration.^[27] Neither the Proposed Rule nor the PRIA cites or summarizes any such research. The PRIA analyzes the residence criteria principally in terms of which foreign citizens would be excluded from apportionment; it does not discuss the tax-record standard, the coverage of the records on which that standard depends, or the standard’s effect on where residents are counted.^[28]

B. Puerto Rico’s address system thwarts reliable matching of records to dwellings.

A records-based residence standard is only as reliable as the link between a record’s address and a housing unit. In Puerto Rico that link is weak, and the Bureau has documented the reasons for that weakness. According to the Bureau’s 2010 assessment, 36 percent of the Puerto Rico Master Address File (“MAF”) units had non-city-style addresses at the start of 2010 Census operations, and 25 percent still did at the end.^[29] Even in the San Juan-Caguas-Guaynabo metropolitan area, only about 77 percent of addresses were classified as city-style.^[30] By comparison, a 2013 Bureau study of administrative-record matching found that non-city-style addresses made up about 3 percent of the roughly 225 million MAF addresses eligible for matching nationwide.^[31] The Bureau has also described the features that weaken matching on the island: rural-route addresses, several houses on a single family lot sharing one address, and cluster-box addresses whose numbers and street names do not identify the location of the associated housing units.^[32]

These features matter because administrative-record matching works almost exclusively on city-style addresses. In the Bureau study just cited, which matched the Bureau’s Statistical Administrative Records System to the MAF, 99 percent of the matched records had an address sourced from the IRS. Of roughly 119 million unit-level matches, only about 3,100 involved non-city-style addresses.^[33] The Bureau’s current address guidelines for Puerto Rico state that it does not attempt to match or add to the MAF any record containing only non-city-style address information.^[34] And the Bureau acknowledges that variant forms of the same Puerto Rico address can generate duplicate MAF records.^[35] Variant forms can arise even for city-style addresses, because a complete Puerto Rico address often includes the name of an urbanization, which the Postal Service treats as “an important part of the address,” but which administrative records may omit or record inconsistently.^[36]

The 2030 Local Update of Census Addresses (LUCA) operation, in which eligible governments, including Puerto Rico’s municipios as county equivalents, may review the address list and suggest additions and corrections between October 2027 and March 2028, is the principal opportunity for local governments to close these gaps before the count. The Bureau’s published description of that operation refers to “descriptive information about the location where people live” and to geographic coordinates, but it does not address Puerto Rico or non-city-style addresses.^[37]

The 2020 Census bears out the risk. Following Hurricane María, the Bureau enumerated all of Puerto Rico through its Update Leave operation.^[38] The Post-Enumeration Survey then estimated a net overcount of 5.7 percent, or about 174,000 people, and erroneous enumerations of 9.8 percent, or about 319,000. Those erroneous enumerations included an estimated 9.1 percent, or about 294,000, duplicate records, and a Bureau official explained that the duplicates went undetected in part because “[w]e didn’t have the same quality of administrative records in Puerto Rico to conduct that unduplication effort.”^[39] The net figure masks errors in both directions: the survey also estimated that the census omitted about 217,000 residents entirely, 7.1 percent of the population the survey estimated. The 2020 net overcount was not statistically different from the 4.5 percent overcount measured in 2010, so the problem predates Hurricane María.^[40]

Puerto Rico’s coverage problem, in other words, was not the number of people counted but records that could not be reconciled with one another or with dwellings: real residents were missed while others were counted twice, and the administrative records needed to reconcile them were weaker on the island than in the States. A tax-record standard keyed to mailing addresses would not cure that problem. It is more likely to compound it, either by leaving out residents whose records cannot be matched without corroborating evidence of residence, or by placing them at a mailing address in a different block, barrio, or municipio. CNE is not aware of any Bureau analysis of how the tax-record standard would perform in Puerto Rico. The available evidence, including the Bureau’s own coverage and address data, indicates that it would increase both omissions and misplacements.

C. The day-count test does not account for residents who divide their time between Puerto Rico and the States.

The tax-record standard is attached to a new residence test. Under proposed § 60.2, a person’s usual residence is the residence at which the person has “lawfully spent the greatest number of days” during a “census enumeration period” running from January 3 to April 1, and a person spends a day at a residence by being “physically and lawfully present there” for at least one hour.^[41] The 2020 criteria, by contrast, counted people at “the place where a person lives and sleeps most of the time.”^[42] The preamble ties the January 3 start date to the congressional term under the Twentieth Amendment, a rationale grounded in congressional representation, and states that the Bureau “intends to consider tax records filed after April 1 insofar as they are probative” of residence during the period.^[43] Beyond stating that “[a] person may spend a certain day at more than one residence,” the Proposed Rule does not address people who divide their time between residences or who move during the period. Unlike the 2020 criteria, which counted people who could not identify where they live most of the time “where they are staying on Census Day,” it provides no fallback rule for ties, and it does not say what records would establish how many days a person spent at a given address.^[44]

That gap is particularly important in Puerto Rico. Migration between the island and the States is large and runs in both directions. In 2024, an estimated 38,000 people moved from Puerto Rico to the States and about 24,000 moved from the States to Puerto Rico, and since 2005 about 68 percent of those moving to the island have been return migrants born there.^[45] Those annual estimates capture only changes of residence. They do not capture residents who move back and forth during the year, a pattern of circular migration that researchers have long documented.^[46] For each such resident, the proposed test would turn on where the person spent more days between January 3 and April 1.

If Part 60 governs the census of Puerto Rico, then that determination would decide where on the island a person is counted. It would also decide whether the person is counted in a State, and therefore in the apportionment base, or in Puerto Rico. The 2020 criteria drew that line expressly, treating Puerto Rico as “outside the United States.”^[47] The Proposed Rule keeps the “stateside census” language of § 60.5(a), which turns on where a person was “living outside the United States on Census Day,” but drops the definition, and it does not say whether a person is “living” outside the United States on Census Day by reference to § 60.2’s count of days or by some other measure.^[48] The records on which the Proposed Rule would rely are of little use in making these determinations on the island: federal returns record addresses and income sources rather than days present, the Proposed Rule describes no access to Hacienda returns (Part II.A), and many island addresses cannot be reliably tied to a dwelling (Part II.B). And because dependents without a consistent separate residence would be counted at whichever of their providers’ usual residences they spent the most days, the same uncertainty would extend to children.^[49]

D. Errors in the base count would propagate through Puerto Rico’s statistics, funding, and representation.

Errors in the decennial base do not remain isolated in the decennial file. PRCS weights are controlled to the Bureau’s official population estimates,^[50] which are carried forward from the decennial base. In Puerto Rico, where migration flows are large (Part II.C), errors in that base compound over the decade, and the PRIA itself acknowledges that without new decennial data to recalibrate those estimates, ACS weights and estimates become less reliable.^[51] Puerto Rico’s official labor statistics rest on the same base. Puerto Rico is outside the Current Population Survey,^[52] which is designed to represent each State and the District of Columbia; its labor force estimates come instead from the Encuesta de Grupo Trabajador, a household survey whose results are controlled each year to population estimates by age and sex that the Puerto Rico Institute of Statistics prepares from the 2020 Census.^[53]

In fiscal year 2021, 353 federal assistance listings used Decennial Census Programs data, in whole or in part, to distribute more than $2.8 trillion.^[54] In Puerto Rico, those data also determine which areas qualify for Community Development Block Grant area-benefit activities, through low- and moderate-income summary data that HUD derives from five-year ACS estimates, which for Puerto Rico are PRCS estimates.^[55] Tract-level income and poverty estimates from the PRCS also determine eligibility for place-based incentives, including the New Markets Tax Credit, HUBZone designations (for which a special Puerto Rico rule expires in December 2027), and, beginning in 2027, qualified opportunity zones. The statutory rule that deemed every low-income census tract in Puerto Rico an opportunity zone is repealed effective December 31, 2026, so Puerto Rico tracts will no longer be designated automatically; they must qualify on PRCS income and poverty estimates and be nominated, as tracts in the States are.^[56] The certified fiscal plan of the Financial Oversight and Management Board for Puerto Rico also builds its population projections on census results, and the projected rate of population change drives its forecasts of several revenue streams, including excise revenues from alcoholic beverages, tobacco, and petroleum.^[57]

The Constitution of Puerto Rico requires that the island’s senatorial and representative districts be reviewed after each decennial census, a review conducted with Census Bureau population figures.^[58] Because the constitutional board that redraws the island’s 8 senatorial and 40 representative districts assigns territory down to the barrio level, households omitted or placed at the wrong address under Part 60 would distort the districts drawn from the 2030 Census, which would govern elections beginning in 2032.^[59] The Proposed Rule’s statement that any effect on governmental entities would be “indirect”^[60] does not account for any of these channels as they operate in Puerto Rico.

CNE therefore respectfully requests that the Bureau:

4. Provide that, in Puerto Rico, the absence of a federal income tax return shall not be treated as evidence against residence.
5. Confirm that Form 499R-2/W-2PR is a “W-2 form” acceptable for purposes of proposed § 60.2(a).
6. State whether it will use Hacienda or other territorial records and, if so, under what authority and agreement and with what safeguards.
7. Before relying on any administrative source in Puerto Rico, publish that source’s coverage and address-match rates for the island, together with the current share of Puerto Rico MAF units with non-city-style addresses.
8. Retain self-response and field enumeration as the primary evidence of residence wherever records cannot be matched to a dwelling.
9. Conduct a Post-Enumeration Survey for Puerto Rico in 2030 and report omissions, erroneous enumerations, and duplicates separately, as it did for 2020.
10. Because proposed § 60.2(a) would make address-matched records evidence of residence, confirm that, in the 2030 LUCA operation, submissions from Puerto Rico that identify housing units by physical description or geographic coordinates will be used to add and locate units, notwithstanding the statement in the Bureau’s Minimum Address Data Submission Guidelines for Puerto Rico that it does not match or add records containing only non-city-style address information.
11. (i) Explain how the day count in proposed § 60.2 would determine whether a person who divides time between Puerto Rico and the States is counted in what proposed § 60.5(a) calls the “stateside census” or in the census of Puerto Rico, including whether a person whose day count favors a Puerto Rico residence is “living outside the United States on Census Day,” given that Part 60 does not define “United States”; (ii) identify the records that would establish days present; and, absent that explanation, (iii) apply a single rule on both sides of that line: the 2020 standard of “the place where a person lives and sleeps most of the time,” with the 2020 fallback to where the person is staying on Census Day.

3. Proposed Part 70 Would Eliminate Puerto Rico’s Only Full-Count Demographic Benchmark Without Considering Puerto Rico

A. Part 70 would change Puerto Rico’s questionnaire without consulting Puerto Rico or weighing the reliance interests built on it.

By its terms, proposed § 70.2(a) bars race and ethnicity questions on “any questionnaire used for the enumeration of the population,” language broad enough to reach Puerto Rico’s questionnaire (see Part I). Before 2000, Puerto Rico’s questionnaire was developed with the Puerto Rico Planning Board, the Government of Puerto Rico’s liaison under a 1958 Memorandum of Agreement with the Bureau; the race question had been dropped from it in 1960, and it did not include the Hispanic-origin question asked in the States. For Census 2000, the Government of Puerto Rico asked the Bureau to use the same questionnaire content as in the States, and since then Puerto Rico’s questionnaire, data processing, and data products have tracked those of the States.^[61]

Part 70 would take away by regulation the race and Hispanic-origin content that Puerto Rico requested, without mentioning that history and without indicating whether the Government of Puerto Rico was consulted. That the prohibition would also apply in the States does not settle the point. Whatever its reasons, Puerto Rico’s request gave the island the same race and Hispanic-origin data the States receive, now spanning three decennial censuses and two decades of PRCS estimates. Losing those data would also cost Puerto Rico more than the States: the island has few other sources of race data, and the surname-based method used elsewhere to estimate race cannot distinguish Puerto Rico’s residents by race (Part III.B). An agency that changes course must show awareness that it is doing so and take serious reliance interests into account.^[62] The Proposed Rule and the PRIA do neither as to Puerto Rico.

B. The 2020 results make a comparable 2030 count indispensable.

Puerto Rico’s race data changed sharply between 2010 and 2020. The share of residents identifying as White alone fell from 75.8 percent to 17.1 percent, and nearly half of the population identified with two or more races.^[63] The Bureau has attributed the nationwide 2010–2020 shifts largely to improvements in question design, data processing, and coding, and has said that comparisons between 2020 and 2010 race data “should be made with caution.”^[64] Others have attributed part of Puerto Rico’s shift to changing self-identification, including campaigns urging residents to embrace their African heritage.^[65] Which explanation holds for Puerto Rico is an open empirical question. A 2030 full count under comparable methods is the only full-count way to answer it. Part 70 would leave Puerto Rico with a single, methodologically discontinuous full-count observation and no full-count means of resolving it. As one scholar of race in Puerto Rico, Yarimar Bonilla, has asked: “Will we see an intensification of this pattern, or will 2020 have been kind of a blip moment?”^[66] The answer matters beyond social science research.

Apart from the decennial census and the PRCS sample, Puerto Rico has few sources of race data; the government does not collect racial data on populations such as people who are homeless or incarcerated, and during the COVID-19 pandemic there was, as the Associated Press reported, “no racial data on coronavirus testing, hospitalizations or fatalities” on the island.^[67] A comparable decennial count is the only full-count benchmark for the size and distribution of Puerto Rico’s population by race. Section 70.2 would also eliminate the Hispanic-origin question, including its “Puerto Rican” response. Although nearly all residents identify as Hispanic, that question is the only full-count source on the composition of the island’s Hispanic population, including the size of its Dominican and other communities, at the municipio and tract level. It is also the only full-count source identifying people of Puerto Rican origin in the States.^[68]

The PRCS cannot fill that role. Proposed § 70.4 preserves race and ethnicity questions on the American Community Survey and similar surveys, but the PRCS samples approximately 36,000 addresses a year, and its person weights are controlled to population estimates by total population, age, and sex, not by race or ethnicity.^[69] Its race and ethnicity estimates for municipios, barrios, and tracts would therefore be sample estimates with no full-count benchmark against which to evaluate them.

The PRIA concedes this point. It acknowledges that decennial population totals for race and ethnicity groups are more reliable than ACS estimates, that ACS one-year estimates are available only for areas of at least 65,000 people, and that dropping race and ethnicity from the 2030 questionnaire would force estimates onto outdated or less accurate data because there are “no other comparably accurate, comprehensive, and reliable sources of this information.”^[70] Only 11 of Puerto Rico’s 78 municipios had 65,000 or more residents in the 2020 Census.^[71] Yet the PRIA’s distributional analysis names only the Island Areas, where the ACS is not conducted on an ongoing basis,^[72] and it proposes consultation only with tribes, on the ground that the ACS cannot produce accurate estimates for small tribal populations.^[73] The PRIA offers no reason why the reasoning that leads the Bureau to consult tribes does not also lead it to consult the Government of Puerto Rico.

The PRIA also identifies, but does not assess, a further category of cost. It acknowledges that the intercensal race and ethnicity estimates built on decennial counts serve as “population controls for major federal surveys” and as “denominators for crucial social and economic indicators,” and that without decennial race and ethnicity data those estimates would rest on “outdated and/or less accurate data.”^[74]

The same benchmarks are also used to test statistical models and automated decision systems, including those built with artificial intelligence, for bias. Testing whether a system treats a group fairly requires knowing who belongs to the group; where records do not say, analysts infer it from census distributions of race and ethnicity by surname and geography. As the National Institute of Standards and Technology has observed, “[d]emographic group membership, necessary for bias testing purposes, is often inferred using the Bayesian improved surname geocoding (BISG) process.”^[75] The Centers for Medicare & Medicaid Services developed and distributes a version of that method for comparing the health care received by Medicare beneficiaries of different racial and ethnic groups, and fair-lending regulators have used it because lenders generally may not ask applicants for non-mortgage credit their race or ethnicity.^[76] Weaker census benchmarks mean a weaker ability to detect discriminatory outcomes. The Bureau “seeks public comment on . . . the various ways in which Census data is utilized across the country,”^[77] yet the PRIA does not address these uses.

The gap is widest in Puerto Rico. The Consumer Financial Protection Bureau’s published version of that method covers only “the contiguous U.S., Alaska, and Hawaii,”^[78] and it treats Hispanic origin as a category separate from race.^[79] Because nearly all of Puerto Rico’s residents identify as Hispanic,^[80] the method cannot distinguish them by race or color. Testing for disparities by race and color on the island therefore relies more heavily on direct population benchmarks, and for municipios, barrios, and tracts the decennial census is the only full-count source of them. Part 70 would remove that benchmark where it is least replaceable.

C. Proposed § 70.3(b) threatens small-area tables and does not say how it applies to Puerto Rico.

Proposed § 70.3(b) provides that the Bureau “shall not disclose or publish any individual demographic data if such disclosure or publication necessitates introducing uncertainty into the population count data, at either the national or the state level.”^[81] The Proposed Rule does not define “individual demographic data.” The preamble acknowledges that in 2020 the Bureau introduced uncertainty into population counts below the state level, and it cites a Department order prohibiting noise infusion.^[82] Because the 2020 masking was applied to published tables, the provision appears aimed at those tables, not only at records about individuals. If counts cannot be perturbed, then confidentiality must be protected by other means, most plausibly by suppressing or coarsening demographic tables. That burden would fall hardest on small geographies, including the municipio, barrio, tract, and block tables used by planners, municipal governments, and disaster-recovery programs in Puerto Rico. The Bureau treated Puerto Rico as a state equivalent for 2010 Census data products,^[83] but proposed § 70.3(b) does not say whether Puerto Rico is “the state level” for this purpose.

D. The PRIA’s own cost-benefit framework weighs against applying Part 70 in Puerto Rico.

The PRIA’s own framework also weighs against applying Part 70 in Puerto Rico. It identifies two core benefits: a more accurate count for apportionment, which has no application in Puerto Rico, and reduced disclosure risk, which it does not evaluate for Puerto Rico.^[84] The costs it describes, by contrast, would fall on Puerto Rico in full. The PRIA’s remaining benefit, reduced respondent burden, is modest (it removes the race and Hispanic-origin items from the short form) and cannot offset the loss of the only full-count race and ethnicity data for Puerto Rico. The only alternative the PRIA considers is no action.^[85] It does not consider applying Part 70 differently in Puerto Rico, although OMB Circular A-4, on which the PRIA relies, directs agencies to consider different requirements for different geographic regions where benefits and costs vary significantly by region.^[86]

CNE therefore respectfully requests that the Bureau:

12. Exclude the Puerto Rico enumeration from proposed § 70.2(a) and retain the race and Hispanic-origin questions on Puerto Rico’s questionnaire; in any event, consult the Government of Puerto Rico, including the Puerto Rico Planning Board and the Puerto Rico Institute of Statistics, before applying any part of Part 70 to Puerto Rico.
13. State whether Puerto Rico is “the state level” under proposed § 70.3(b), and publish the disclosure avoidance approach it would apply to Puerto Rico’s small-area tables.
14. Explain how comparable municipio-, tract-, and block-level race and ethnicity data for Puerto Rico would be produced after 2030.
15. Assess, in the final regulatory impact analysis, the cost of degraded race and ethnicity benchmarks for statistical models and automated decision systems, including the methods used to test those systems for bias, and the consequences for Puerto Rico.

Conclusion

The Proposed Rule would change how more than three million residents of Puerto Rico are counted, where they are placed, and what is known about them, yet it is defective in each of the three respects identified at the outset. First, it never states whether or how Parts 60 and 70 govern the census of Puerto Rico (requests (1)–(3)). Second, its proposed residence standard rests on federal tax records that many residents of Puerto Rico are not required to file; on addresses that often cannot be matched to a dwelling; and on a day count that does not account for residents who divide their time between the island and the States (requests (4)–(11)). Third, Part 70 would eliminate Puerto Rico’s only full-count race and ethnicity benchmark, and could thin its small-area tables, without any analysis of those losses and without any indication that the Government of Puerto Rico was consulted (requests (12)–(15)).

A final rule adopted on this record could govern the census of Puerto Rico without any analysis of its effects there. CNE therefore respectfully requests the Bureau to withdraw the Proposed Rule or, at a minimum, to re-propose it with an express statement of its application to Puerto Rico, an analysis of its effects there, and consultation with the Government of Puerto Rico. CNE would welcome the opportunity to provide additional information and can be reached at the contact information below.

Respectfully submitted,

/s/Sergio M. Marxuach
Policy Director and General Counsel
Center for a New Economy

 

Endnotes

[1] Decennial Census of the Population of Americans; Proposed Residence Criteria and Proposed Regulations for Demographic Questions, 91 Fed. Reg. 57,524 (proposed Sept. 10, 2026) (to be codified at 15 C.F.R. pts. 60, 70) [hereinafter Proposed Rule].

[2] U.S. Census Bureau, Preliminary Regulatory Impact Analysis: Decennial Census of the Population of Americans; Proposed Residence Criteria and Proposed Regulations for Demographic Questions, Docket No. 260903-0005 (Docket ID USBC-2026-0628-0002) (posted Sept. 10, 2026), https://www.regulations.gov/document/USBC-2026-0628-0002 [hereinafter PRIA].

[3] 13 U.S.C. § 191(a).

[4] U.S. Census Bureau, 2010 Census of Puerto Rico Assessment Report, 2010 Census Planning Memoranda Series No. 238, at 2 (Sept. 18, 2012) [hereinafter 2010 Puerto Rico Assessment].

[5] Proposed Rule, 91 Fed. Reg. at 57,529 (authority for proposed Part 60: 13 U.S.C. §§ 4, 141); id. at 57,530 (authority for proposed Part 70: 13 U.S.C. §§ 4, 5, 141).

[6] Id. at 57,524.

[7] Proposed 15 C.F.R. § 60.1, 91 Fed. Reg. at 57,529.

[8] Proposed 15 C.F.R. § 70.1, 91 Fed. Reg. at 57,530.

[9] Proposed Rule, 91 Fed. Reg. at 57,524 (summary); proposed 15 C.F.R. § 70.2(a), 91 Fed. Reg. at 57,530.

[10] Proposed 15 C.F.R. § 70.3(a), 91 Fed. Reg. at 57,530.

[11] P.R. Const. art. III, § 4.

[12] Final 2020 Census Residence Criteria and Residence Situations, 83 Fed. Reg. 5525, 5526 n.1, 5533 n.9 (Feb. 8, 2018).

[13] Proposed 15 C.F.R. § 60.5(a), 91 Fed. Reg. at 57,529–30.

[14] Proposed Rule, 91 Fed. Reg. at 57,528.

[15] PRIA, supra note 2, at 12, 14 & n.12.

[16] 13 U.S.C. § 191(a)–(b).

[17] PRIA, supra note 2, at 14 n.13.

[18] Proposed 15 C.F.R. § 60.2(a), 91 Fed. Reg. at 57,529; Proposed Rule, 91 Fed. Reg. at 57,528 n.7.

[19] 26 U.S.C. § 933(1).

[20] Internal Revenue Serv., Publ’n 1321, Special Instructions for Bona Fide Residents of Puerto Rico Who Must File a U.S. Individual Income Tax Return (rev. Oct. 2025).

[21] Internal Revenue Serv., More Puerto Rico Families Qualify for Child Tax Credit (May 16, 2022), https://www.irs.gov/newsroom/more-puerto-rico-families-qualify-for-child-tax-credit.

[22] Proposed Rule, 91 Fed. Reg. at 57,529 n.17.

[23] Id. at 57,526.

[24] 13 U.S.C. § 6(b).

[25] Internal Revenue Serv., Form W-3 (PR), Transmittal of Withholding Statements (2026) (directing employers to send Forms 499R-2/W-2PR to the Social Security Administration).

[26] Segundo Castro-González & Julio C. Medina León, Economic Crisis’s Influence over the Quantification of Puerto Rico’s Shadow Economy: Parsimonious Econometric Proposal, 19 J. Applied Bus. & Econ., no. 8, 2017, at 59 (estimating Puerto Rico’s shadow economy at 36.47% of GDP in 2014); José Caraballo-Cueto, Informal Labor Markets After Disasters: Evidence from Rural and Urban Low-Income Communities in Puerto Rico, 107 Soc. Sci. Q. e70148 (2026).

[27] 2010 Puerto Rico Assessment, supra note 4, at xii.

[28] PRIA, supra note 2, at 11–13.

[29] 2010 Puerto Rico Assessment, supra note 4, at xi. These figures count Master Address File units with non-city-style addresses. They do not capture city-style addresses that fail to match because of local address formats, and local estimates of structures lacking an address that federal systems can validate are higher. See, e.g., Puerto Rico Cuenta, Impacto en PR, https://prcuenta.com/impacto (last visited Oct. 21, 2026) (citing estimates by iCasaPR that 40 to 60 percent of structures lack such an address).

[30] 2010 Puerto Rico Assessment, supra note 4, at xii.

[31] Christine Gibson Tomaszewski, Updating the U.S. Census Bureau’s Master Address File via Administrative Records, in JSM Proceedings, Government Statistics Section 3638, 3640 tbl.1 (Am. Statistical Ass’n 2013).

[32] 2010 Puerto Rico Assessment, supra note 4, at 31.

[33] Tomaszewski, supra note 31, at 3641 tbl.3, 3643.

[34] U.S. Census Bureau, Minimum Address Data Submission Guidelines – Puerto Rico, https://www2.census.gov/geo/pdfs/partnerships/data_guidelines/Minimum_Address_Guidelines_PR.pdf (last visited Oct. 16, 2026).

[35] U.S. Census Bureau, Puerto Rico Community Survey: Accuracy of the Data (2024), at 23 [hereinafter PRCS Accuracy].

[36] U.S. Postal Serv., Publ’n 28, Postal Addressing Standards § 295.5, https://pe.usps.com/text/pub28/28c2_045.htm (last visited Oct. 21, 2026).

[37] U.S. Census Bureau, Local Update of Census Addresses (LUCA) Operation, https://www.census.gov/programs-surveys/decennial-census/about/luca.2030.html (last updated Aug. 12, 2026).

[38] U.S. Census Bureau, 2020 Census Plans for the Commonwealth of Puerto Rico, 2020 Census Memorandum Series No. 2018.13 (2018).

[39] Press Release, U.S. Census Bureau, Census Bureau Releases Results of Post-Enumeration Survey for Puerto Rico From 2020 Census (Aug. 16, 2022) (reporting an estimated 9.1% or 294,000 duplicate records); Nicole Acevedo, 2020 Census Overcounted People in Puerto Rico, Hinting at Larger Population Loss, NBC News (Aug. 17, 2022) (quoting Timothy Kennel, U.S. Census Bureau).

[40] U.S. Census Bureau, 2020 Census Post-Enumeration Survey in Puerto Rico slides 11–15 (Sept. 2022), https://www.census.gov/content/dam/Census/newsroom/press-kits/2022/09142022-presentation-2020-pes-pr-english.pdf (estimating 217,000 omissions, 7.1% of the survey’s population estimate); Press Release, U.S. Census Bureau, supra note 39 (reporting that the 2020 overcount was not statistically different from the 4.5% overcount in 2010); see also Fin. Oversight & Mgmt. Bd. for P.R., Second Revised 2024 Fiscal Plan for Puerto Rico 36 (certified June 19, 2026) [hereinafter Fiscal Plan] (attributing Puerto Rico’s historical overcounts to “challenges linked to Puerto Rico’s administrative records, which make de-duplication of responses more difficult than on the U.S. mainland”).

[41] Proposed 15 C.F.R. § 60.2(a)–(c), 91 Fed. Reg. at 57,529.

[42] Final 2020 Census Residence Criteria and Residence Situations, 83 Fed. Reg. at 5526.

[43] Proposed Rule, 91 Fed. Reg. at 57,526.

[44] Proposed 15 C.F.R. § 60.2(b), 91 Fed. Reg. at 57,529; Final 2020 Census Residence Criteria and Residence Situations, 83 Fed. Reg. at 5533.

[45] Inst. de Estadísticas de P.R., Perfil del Migrante 2023-2024, at 5–6, 15, 39 tbl.A1 (Apr. 29, 2026) (estimates from the American Community Survey and the Puerto Rico Community Survey).

[46] See Jorge Duany, Mobile Livelihoods: The Sociocultural Practices of Circular Migrants Between Puerto Rico and the United States, 36 Int’l Migration Rev. 355 (2002).

[47] Final 2020 Census Residence Criteria and Residence Situations, 83 Fed. Reg. at 5533 n.9.

[48] Proposed 15 C.F.R. § 60.5(a), 91 Fed. Reg. at 57,529–30.

[49] Proposed 15 C.F.R. § 60.3(a)–(b), 91 Fed. Reg. at 57,529.

[50] PRCS Accuracy, supra note 35, at 11, 15.

[51] PRIA, supra note 2, at 12–13.

[52] U.S. Bureau of Lab. Stat., Effects of Hurricanes Harvey, Irma, and Maria on BLS Data Collection and Reporting, https://www.bls.gov/bls/hurricanes-harvey-irma-maria.htm (last updated Jan. 30, 2019) (“Puerto Rico conducts its own monthly household survey, patterned after the Current Population Survey, to produce labor force information.”).

[53] U.S. Bureau of Lab. Stat., Current Population Survey, https://www.bls.gov/cex/cecomparison/cps_profile.htm (last modified Jan. 24, 2023) (describing a sample designed “to represent each state and the District of Columbia”); P.R. Dep’t of Lab. & Hum. Res., Empleo y Desempleo en Puerto Rico (Encuesta de Grupo Trabajador): Agosto 2026 (Sept. 18, 2026), https://mercadolaboral.pr.gov/lmi/pdf/Default/Grupo%20Trabajador/EMPLEO%20Y%20DESEMPLEO%20EN%20PUERTO%20RICO.pdf.

[54] Ceci Villa Ross, U.S. Census Bureau, Uses of Decennial Census Programs Data in Federal Funds Distribution: Fiscal Year 2021 (Working Paper, June 2023).

[55] U.S. Dep’t of Hous. & Urb. Dev., Notice CPD-24-04, Low- and Moderate-Income Summary Data Updates (June 6, 2024) (describing summary data based on 2016–2020 ACS five-year estimates, except for American Samoa, Guam, the Northern Mariana Islands, and the U.S. Virgin Islands, which use the 2020 Island Areas Census); U.S. Census Bureau, Puerto Rico Community Survey, https://www.census.gov/programs-surveys/acs/about/puerto-rico-community-survey.html (last visited Oct. 21, 2026) (describing the PRCS as “part of the Census Bureau’s American Community Survey, customized for Puerto Rico”).

[56] Cmty. Dev. Fin. Insts. Fund, U.S. Dep’t of the Treasury, CDFI Fund to Release Updated Program Eligibility Information: 2016-2020 American Community Survey Census Bureau Data (June 30, 2022); Off. of Tax Analysis, U.S. Dep’t of the Treasury, Eligibility Criteria for Determining Census Tracts Eligible to be Nominated as 2027 Qualified Opportunity Zones (Mar. 2026) (applying 2020–2024 ACS five-year estimates to tracts in the 50 States, the District of Columbia, and Puerto Rico); 26 U.S.C. § 1400Z-1(b)(3), repealed by Pub. L. No. 119-21, § 70421(a)(3), (a)(5)(B), 139 Stat. 72, 223–24 (2025) (effective Dec. 31, 2026); 13 C.F.R. § 126.103 (2026) (defining HUBZones to include qualified census tracts, which HUD designates using census household income and poverty data, and suspending the 20 percent population limit for Puerto Rico’s metropolitan areas until December 22, 2027); HUBZone and Puerto Rico Oversight, Management, and Economic Stability Act (PROMESA) Amendments, 82 Fed. Reg. 48,903 (Oct. 23, 2017) (implementing PROMESA § 412(a)).

[57] Fiscal Plan, supra note 40, at 36–37 & exh. 20.

[58] P.R. Const. art. III, § 4; Junta Constitucional de Revisión de los Distritos Senatoriales y Representativos de Puerto Rico, Principios para la Revisión de los Distritos Senatoriales y Representativos (Nov. 16, 2021).

[59] Así quedó el nuevo mapa electoral de distritos senatoriales y representativos, Noticel (July 21, 2022) (describing the map, in effect for the 2024 and 2028 elections, and its assignment of individual barrios to districts).

[60] Proposed Rule, 91 Fed. Reg. at 57,528.

[61] 2010 Puerto Rico Assessment, supra note 4, at 2–3, 8; Dánica Coto, Puerto Rico Ponders Race Amid Surprising Census Results, Associated Press (Oct. 15, 2021) (reporting that “[f]rom 1960 to 2000, the island conducted its own census and never asked about race”).

[62] FCC v. Fox Television Stations, Inc., 556 U.S. 502, 515 (2009); Encino Motorcars, LLC v. Navarro, 579 U.S. 211, 221–22 (2016).

[63] Damayra Figueroa-Lazu, Jennifer Hinojosa & Yarimar Bonilla, Ctr. for Puerto Rican Studies, Hunter Coll., Puerto Rico’s 2020 Race/Ethnicity Decennial Analysis (2022) (analyzing 2010 and 2020 Census redistricting data); see also U.S. Census Bureau, 2020 Census Redistricting Data (Pub. L. 94-171), tbl. P1 (Puerto Rico).

[64] Nicholas Jones, Rachel Marks, Roberto Ramirez & Merarys Ríos-Vargas, U.S. Census Bureau, 2020 Census Illuminates Racial and Ethnic Composition of the Country (Aug. 12, 2021).

[65] Coto, supra note 61 (reporting that “grass-roots organizations in Puerto Rico launched campaigns to urge people to embrace their African heritage”).

[66] Id. (quoting Yarimar Bonilla).

[67] Id. (reporting the absence of racial data on COVID-19 outcomes).

[68] U.S. Census Bureau, 2020 Census Informational Questionnaire question 8, https://www2.census.gov/programs-surveys/decennial/2020/technical-documentation/questionnaires-and-instructions/questionnaires/2020-informational-questionnaire-english_DI-Q1.pdf (Hispanic-origin question, including “Yes, Puerto Rican”); see supra note 61 and accompanying text.

[69] PRCS Accuracy, supra note 35, at 6, 16.

[70] PRIA, supra note 2, at 12–15.

[71] U.S. Census Bureau, 2020 Census Redistricting Data (Pub. L. 94-171), tbl. P1 (Puerto Rico municipios).

[72] PRIA, supra note 2, at 17.

[73] Id. at 10, 14–15.

[74] Id. at 15.

[75] Reva Schwartz et al., Nat’l Inst. of Standards & Tech., Special Publ’n 1270, Towards a Standard for Identifying and Managing Bias in Artificial Intelligence 30 (Mar. 2022), https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.Sp.1270.pdf.

[76] Ctrs. for Medicare & Medicaid Servs., Chronic Conditions Warehouse, Medicare Bayesian Improved Surname Geocoding (MBISG) File User Guide §§ 2.0, 3.4 (version 1.4, June 2026); Research Data Assistance Ctr., Medicare Bayesian Improved Surname Geocoding (MBISG) Data Now Available (Mar. 27, 2024) (describing MBISG as “developed by CMS”); Consumer Fin. Prot. Bureau, Using Publicly Available Information to Proxy for Unidentified Race and Ethnicity 3–4, 7–8 (Summer 2014), https://files.consumerfinance.gov/f/201409_cfpb_report_proxy-methodology.pdf; 12 C.F.R. § 1002.5(b).

[77] PRIA, supra note 2, at 9.

[78] Consumer Fin. Prot. Bureau, supra note 76, at 8 n.15.

[79] Id. at 7 & n.9.

[80] U.S. Census Bureau, 2020 Census Redistricting Data (Pub. L. 94-171), tbl. P2, Hispanic or Latino, and Not Hispanic or Latino by Race (Puerto Rico) (98.9% Hispanic or Latino).

[81] Proposed 15 C.F.R. § 70.3(b), 91 Fed. Reg. at 57,530.

[82] Proposed Rule, 91 Fed. Reg. at 57,527; id. at 57,529 n.13 (citing Department Administrative Order 216-26); see U.S. Dep’t of Com., Dep’t Admin. Order 216-26, Disclosure Avoidance for Statistical Products (effective June 4, 2026).

[83] 2010 Puerto Rico Assessment, supra note 4, at ix, 28.

[84] PRIA, supra note 2, at 16.

[85] Id. at 16–17.

[86] Off. of Mgmt. & Budget, Exec. Off. of the President, Circular A-4, Regulatory Analysis 8 (Sept. 17, 2003); see PRIA, supra note 2, at 6 n.1.