What is collected, and what is released

Data Asks

Every other page on this site reports what the public record says. This one reports where it stops.

Each ask below names something a federal office holds, or once published, that would answer a question this site currently cannot. It says what is missing, and it says which analysis is blocked. The figures beside each ask are read from this site's own built data when the page is rendered, so an ask cannot outlive the evidence for it.

16 of 24 are not requests for new collection. The agency already has the field on a form it already processes, or published it until recently and stopped. Those are disclosure decisions. The rest are marked, because a release decision and a budget request are different conversations.

Nothing here asks for anything that identifies a person. Several asks are deliberately for a tabulation rather than for record-level data, because the tabulation answers the question and the records would not be released.

asks

across 10 departments and agencies

already collected

on a form the agency already processes

published, then stopped

the collection did not change; the release did

would need new collection

a cost, and a different conversation

How to read an ask

The badge is the part that decides who you are talking to.

Published before, then stopped

This was public and is not any more. The collection has not changed; only the release has.

No new collection

The agency holds this today, on a form or survey it already processes. Publishing it is a disclosure decision.

A tabulation would do it

The record-level data is sensitive or suppressed, but a tabulation at the agency's own thresholds would answer the question without releasing it.

Needs a new collection

Nobody holds this. It is a real cost and belongs in a different conversation from the rest of this page.

The asks

Grouped by the office that holds the data.

USCIS

1. USCIS

H-1B cap-exempt status by employer

No new collection

Collected on: Form I-129, which records whether a petition is cap-exempt and on what ground.
Published now: National totals by exemption category in the Report to Congress. The H-1B Employer Data Hub names every employer and carries no cap field.
The request: The Employer Data Hub extract with two added fields, cap-exempt yes or no and the exemption ground, by employer and fiscal year, with approvals and denials.
Would settle: Which employers use the exemption, at what scale, and whether use is concentrated. Today the petitions filed under the affiliated-nonprofit ground cannot be attributed to any organisation.

2,388employers on a cap-exempt list built by inference, not by record

Built at render time from cap_exempt_directory.json:captured_total.

6. USCIS

L-1 and O-1 petitions by employer

No new collection

Collected on: Form I-129, the same form and system that produce the published H-1B Employer Data Hub.
Published now: Nothing by employer. Neither classification requires a labour condition application, so the Labor Department never sees them, and USCIS publishes employer hubs for H-1B, H-2A and H-2B only.
The request: Petitioner name and address for L-1A, L-1B, O-1A and O-1B, aggregated by petitioner and fiscal year, with approvals and denials. The H-1B hub's extract, run for four more classifications.
Would settle: Which companies transfer the most staff into the United States without a wage test, and whether O-1 growth sits in the industries most constrained by the H-1B cap.

10. USCIS

A stable employer identifier in the H-1B Employer Data Hub

No new collection

Collected on: Form I-129, which carries the petitioner's FEIN.
Published now: The employer name as filed, and nothing else. There is no key that survives a spelling change.
The request: Any stable per-employer identifier alongside the name. A hashed or truncated FEIN would do. The agency need not publish the number itself, only something that is the same across filings by the same body.
Would settle: Counting an employer at all. Cognizant files under eighteen spellings and the largest accounts for 22,287 of 50,994. Every employer total published anywhere, by anyone, is a floor, and the error is largest for exactly the biggest filers.

11. USCIS

Name the affiliated nonprofits in the Report to Congress

No new collection

Collected on: Form I-129.
Published now: A count of petitions from nonprofits related to or affiliated with an institution of higher education. No organisation is named.
The request: The same count, broken out by petitioning organisation.
Would settle: Sizing the bodies running the cap-exempt route. This site reaches for IRS Form 990 to do it, which gives revenue and headcount but says nothing about cap status, so the two can sit side by side and never join.

Department of Labor

9. Department of Labor, Office of Foreign Labor Certification

Restore the worker fields to the PERM disclosure file

Published before, then stopped

Collected on: Form ETA-9089, on which the employer still answers all three questions.
Published now: Nothing. Country of citizenship, country of birth and class of admission were all published through FY2024 and all three were removed at once in FY2025, when the file went from 155 columns to 137.
The request: The three withdrawn columns, restored to the quarterly disclosure file on the basis used through FY2024.
Would settle: Every question about who fills a sponsored post. This site can say that a Washington state agency sponsored a Hong Kong citizen for permanent residence in FY2021. It cannot say whether that still happens, and no future harvest will tell us, because the field is gone rather than empty.
Note: This is the single most consequential item on the page. The collection did not change. Only the release did.

38,705filings the ended series covers

FY2013-FY2024window, and it cannot be extended

Built at render time from public_employer_perm.json:public_rows, public_employer_perm.json:window.

12. Department of Labor, Office of Foreign Labor Certification

A six-digit NAICS code on the labour condition application

No new collection

Collected on: Form ETA-9035, on which the employer reports NAICS.
Published now: Whatever depth the employer chose. The same industry appears at two, four, five and six digits in a single file.
The request: Validation requiring six digits at filing, and the full six-digit code in the disclosure file.
Would settle: Separating a staffing firm from a law firm, which is the whole basis of the body-shop test. A filing coded '6111' is a school and cannot be counted as one, because '6111' covers several six-digit industries and assigning it would be a guess.

8,081,883certified applications read

4,783distinct NAICS codes, at mixed depths

Built at render time from h1b_by_industry.json:certified_filings, h1b_by_industry.json:distinct_codes.

13. Department of Labor, Office of Foreign Labor Certification

A published column schema, and a diff when it changes

No new collection

Collected on: Not applicable. This is about the release, not the form.
Published now: Files whose column names change between years without notice. H_1B_DEPENDENT became H-1B_DEPENDENT in FY2025. FW_INFO_BIRTH_COUNTRY became FOREIGN_WORKER_BIRTH_COUNTRY in FY2020.
The request: A versioned schema published with each release, and a diff against the previous one.
Would settle: A whole category of silent error. A renamed column read by its old name returns no value, and a missing dependency flag reads as a firm that is never dependent. That is a wrong answer rather than a crash, and no downstream check can catch it reliably.
Note: Not a request for data. It would cost close to nothing and it is the cheapest item here.

14. Department of Labor, Wage and Hour Division

The complete H-1B enforcement case set

A tabulation would do it

Collected on: WHISARD case management.
Published now: An extract carrying the largest cases only.
The request: All concluded H-1B cases with employer, fiscal year, violation type and back wages, at the division's usual thresholds.
Would settle: Reading absence. A firm that does not appear may have had no findings or may have had small ones, and the extract cannot distinguish them. Only a handful of employers match between the enforcement file and the wage data, so the crossing is inconclusive.

37employers matched across both files

Built at render time from employer_profiles.json:with_enforcement.

DHS

4. DHS Student and Exchange Visitor Program

OPT and STEM OPT for every institution

No new collection

Collected on: SEVIS.
Published now: A Top 100 ranking of schools. A school outside it is absent rather than zero, so school totals are a floor and year-to-year movement may only reflect the cutoff moving.
The request: OPT and STEM OPT authorisation counts for every certified school, by calendar year, on the same basis as the published national table.
Would settle: Work-authorisation participation for every institution, not just the hundred largest.

7. DHS Student and Exchange Visitor Program; the State Department for visa issuance

F-1 students by institution

No new collection

Collected on: SEVIS, which records every F-1 student against a certified school.
Published now: State reports issuance by consulate and nationality, never by school. SEVP publishes school-level figures for OPT alone. IPEDS reports non-resident enrolment, a residency category rather than a visa count.
The request: Active F-1 student records by certified school and year, split by education level.
Would settle: Where the student pipeline enters. That is the first stage of everything else on this page.

16. DHS Office of Immigration Statistics

Country of birth crossed against country of residence

A tabulation would do it

Collected on: Admission and adjustment records.
Published now: Yearbook tables 2 and 3, which are separate totals for the same population.
The request: The cross-tabulation of the two, at the usual thresholds.
Would settle: Whether a surplus in one country can be attributed to a particular origin. No individual is traced across the two tables, so however strongly the deficits in one line up with the surpluses in another, the inference cannot be made, and this site does not make it.

15. DHS SEVP holds the authorisations; Treasury holds the payroll

A wage measure for OPT employment

Needs a new collection

Collected on: SEVIS records the authorisation. No federal collection records what the job pays.
Published now: Nothing.
The request: Aggregate earnings for OPT-authorised employment, by year and broad field. A tabulation from the employer side of the payroll tax would do it without a new survey.
Would settle: Sizing the payroll-tax differential. OPT employment is exempt from the employer share of FICA, which is a real price difference against an equivalent citizen hire. This site publishes no national total for it, only a per-$100,000 figure and a scenario grid, because the wage base does not exist.

NSF

2. NSF National Center for Science and Engineering Statistics

Federal graduate support by citizenship

No new collection

Collected on: Survey of Graduate Students and Postdoctorates, which records support by mechanism and agency and enrolment by citizenship. Survey of Earned Doctorates, which records primary support by citizenship.
Published now: Support by agency, and enrolment by citizenship, never together. The SED records the assistantship but not which award paid for it.
The request: A tabulation of graduate students with federal support, by citizenship status and by field, from existing returns. If the returns do not link the two, a statement saying so, which would itself settle whether the figure exists.
Would settle: What federal research funding spends per doctoral student by citizenship. It would turn a proxy into a measurement.

3. NSF National Center for Science and Engineering Statistics

Citizens and permanent residents separated at conferral

No new collection

Collected on: Survey of Earned Doctorates, which records citizenship status at conferral in three categories.
Published now: Two categories: citizens and permanent residents combined, and temporary visa holders. The Survey of Doctorate Recipients separates them after graduation but not at the degree.
The request: SED doctorate counts by field and year with US citizens and permanent residents shown separately.
Would settle: The citizen share of US graduate STEM doctorates, which no published source reports.
Note: Master's degrees would need a change to IPEDS and are a separate ask, below.

17. NSF National Center for Science and Engineering Statistics

State geography on the Survey of Doctorate Recipients

A tabulation would do it

Collected on: The SDR questionnaire, which records location.
Published now: Nothing below national. State is suppressed in the public-use file for confidentiality.
The request: A state-level tabulation of the core SDR measures with the usual cell suppression. Not the microdata.
Would settle: Every state-level question about the doctoral workforce. This site publishes a page for each of the fifty states and cannot put a single SDR figure on any of them.

18. NSF National Center for Science and Engineering Statistics

Hours worked, above the topcode

A tabulation would do it

Collected on: The SDR questionnaire.
Published now: HRSWKP topcoded at more than 40, so 41 hours and 80 hours are a single value.
The request: Either a higher topcode or a banded distribution above forty.
Would settle: The claim people actually make. The argument is about the distance above forty, and the variable collapses exactly the range the argument is about.

National Center for Education Statistics

20. National Center for Education Statistics

Graduate borrowing in Student Financial Aid

No new collection

Collected on: The IPEDS Student Financial Aid component, which covers undergraduates.
Published now: Undergraduate Pell and federal loans. Graduate borrowing sits in the Education Department's loan systems and not in IPEDS.
The request: Graduate federal loan volume by institution and year in Student Financial Aid, or a crosswalk to the loan data.
Would settle: Putting a federal dollar figure beside a graduate programme. A graduate-heavy institution's federal money is understated on this site, and the borrowing that funds a terminal master's is invisible.

19. National Center for Education Statistics

Permanent residents separated from citizens in IPEDS enrolment

Needs a new collection

Collected on: The IPEDS Fall Enrolment component, which collects 'non-resident alien' as a residency category.
Published now: Two groups, with permanent residents counted as domestic and reported under race and ethnicity.
The request: A permanent-resident category in Fall Enrolment and Completions.
Would settle: Measuring the foreign-born share of a programme. Every non-resident share on this site is a floor, and the gap is largest in exactly the fields where the question is asked most.
Note: This is a change to the collection, not only to the release, which is why it is marked as new. The doctorate version of the same question needs no new collection and is above.

State Department

8. State Department for visa issuance; DHS for admissions

Country of birth for nonimmigrant classes

No new collection

Collected on: The nonimmigrant visa application, which records place of birth alongside nationality.
Published now: Every visa and admission series is by nationality. A naturalised citizen of one country born in another cannot be distinguished from a native-born citizen anywhere in the published data.
The request: Nonimmigrant visa issuances by visa class, fiscal year, nationality and country of birth. For classes admitted without a visa, the equivalent from DHS admission records if they carry birthplace.
Would settle: How far a nationality series reflects the country people actually come from. In the one dataset that publishes both fields, the two disagree often enough to matter, and the largest single citizenship among the mismatches is Canada.

2,056filings where citizenship and birth country differ, where both are published

Built at render time from public_employer_perm.json:anomaly_totals.citizenship_differs_from_birth_country.

Bureau of Labor Statistics

21. Bureau of Labor Statistics

Raise or remove the OEWS wage topcode

A tabulation would do it

Collected on: The OEWS survey.
Published now: Wages topcoded at $239,200 and published as '#', with suppressed cells as '*'.
The request: A higher topcode, or the mean above it, for occupations where the topcode binds.
Would settle: Comparing a high-wage occupation against its prevailing wage. A topcoded occupation has no usable 90th percentile, so the comparison this site runs on every certified filing drops out for the occupations where the money is.

565,966filings priced against the local median

78.7%of those, paid below the local median for the occupation

Built at render time from local_wage_comparison.json:filings_priced, local_wage_comparison.json:below_local_median_pct.

Department of Justice

22. Department of Justice, National Security Division

Resume the research-security case compilation

Published before, then stopped

Collected on: Case records.
Published now: Nothing after 19 November 2021. The compilation stopped there and the China Initiative ended on 23 February 2022.
The request: Resumed publication on the previous basis, or a statement that it has been discontinued.
Would settle: Saying anything about the present. This site publishes the register as a frozen record and labels it as one, because the alternative is to let a 2021 snapshot read as a current picture. Cases have been brought since and none is in a public list.

58entries in the frozen register

7of them brought after the list stopped

Built at render time from security_cases.json:totals.entries, security_cases.json:totals.cases_since_list_stopped.

Internal Revenue Service

23. Internal Revenue Service, Tax Exempt and Government Entities

Restore the Form 990 e-file years that came off irs.gov

Published before, then stopped

Collected on: Form 990, filed electronically.
Published now: Tax years 2019 onward on irs.gov. The 2009 to 2018 e-file XML sat on an AWS public bucket that stopped updating at the end of 2021, and outside researchers report that filings did not survive the move.
The request: The full e-file XML archive back to the start of electronic filing, hosted by the agency, with an index of what is held.
Would settle: Any long-run analysis of a nonprofit. A sponsoring organisation's growth over fifteen years is the question worth asking, and the public record of it now depends on third-party mirrors.
Note: The counts of what went missing come from outside researchers, not from the IRS, and the agency has published no reconciliation. A statement of what is held and what is not would itself be useful.

24. Internal Revenue Service, Tax Exempt and Government Entities

Form 990 employee counts that are populated

No new collection

Collected on: Form 990 Part V, which carries the W-3 employee count.
Published now: The field, frequently as zero even where the filing reports staff elsewhere.
The request: Either a populated field or a documented explanation of when it is left blank.
Would settle: Sizing an organisation by people rather than by money. Revenue is the only usable measure of a cap-exempt sponsor today, and revenue is not a count of anything.

No federal agency

5. No federal agency. The natural home is the National Center for Education Statistics.

Graduate applications and admissions

Needs a new collection

Collected on: Nothing. The IPEDS Admissions component covers first-time undergraduates only.
Published now: Degrees conferred. No federal collection records who applied to a graduate programme, who was admitted, or who was offered funding.
The request: A graduate admissions component in IPEDS: applications, admissions and funded offers by broad field and residency status.
Would settle: Whether domestic applicants are declining, being declined, or not applying. That is the question behind this whole project and the one it cannot answer.

How this page is built

The asks are declared once, in scripts/data_asks.py. Each one names the dataset that demonstrates the gap, and the builder reads the figure from that dataset when the page is rendered.

No figure on this page is typed. An ask whose evidence stops resolving fails the build, so the page cannot keep making a claim the data no longer supports.
Every gap described here was met while building something else on this site. None was looked for.
An ask is about publication, not about collection, unless it says otherwise. Where a field is already on a federal form, the page says so.
Nothing here is a criticism of the staff who publish these files. Most of these gaps are decisions taken years ago for reasons that were good at the time, and several would cost almost nothing to revisit.