Public validation inventory

One of the strongest parts of this proposal is that the model can be validated extensively against public sources before any restricted administrative linkage is available. The public record is fragmented, but it is much richer than many policy-model proposals acknowledge. That matters for two reasons.

First, it means the project can set milestone-based validation gates using sources that outside reviewers can inspect themselves. Second, it reduces the risk that the project becomes blocked on restricted-data negotiations before it can produce a credible first model.

This chapter inventories the main public or low-friction sources we can use to validate the model. It is not an exhaustive bibliography. It is the minimum practical source stack for judging whether longitudinal populace is becoming decision-useful.

Many of these sources are already assembled inside populace, PolicyEngine’s microdata stack — the primary-source microdata and the calibration targets (from CBO, IRS, SSA, Census, and others) that it draws on. Naming the sources explicitly here keeps the validation record legible to outside reviewers, who can inspect the underlying public tables and microdata themselves rather than taking the data layer on trust.

Access tiers

The validation stack naturally breaks into four tiers.

• Open public tables and documentation: SSA statistical publications, Trustees material, Census product pages, CDC mortality tables, CMS public program pages, and state Medicaid documentation.
• Public microdata: CPS ASEC, ACS PUMS, SIPP, PSID public files, HRS public release, NHATS/NSOC public files, and MCBS public-use files (U.S. Census Bureau 2025, 2026, 2024; Panel Study of Income Dynamics 2025; University of Michigan, Institute for Social Research 2025; National Health and Aging Trends Study 2025; Centers for Medicare & Medicaid Services 2025a).
• Low-friction or application-based research files: MCBS limited data, T-MSIS Analytic Files, and some CMS or SSA products available under use agreements rather than full public release (Centers for Medicare & Medicaid Services 2025a, 2025d).
• Restricted administrative linkage: SSA administrative microdata or matched survey-admin files. These could improve precision later, but they are not assumed in the first build.

The proposal should be explicit that stage gates are built primarily on the first two tiers. The third tier is a bonus, not a prerequisite.

Core source inventory

Source	Access tier	Best use in this project	Main caveat
SSA Annual Statistical Supplement (Social Security Administration, Office of Research, Evaluation, and Statistics 2025b)	Open public tables	National benchmarks for insured status, benefit type, awards, terminations, dual entitlement, and SSI distributions	Aggregate, not person-level
SSA DI Annual Statistical Report (Social Security Administration, Office of Research, Evaluation, and Statistics 2025a)	Open public tables	Disability incidence, beneficiary composition, awards, application outcomes, and return-to-work patterns	Administrative program view, not a full longitudinal panel
SSA SSI Annual Statistical Report (Social Security Administration, Office of Research, Evaluation, and Statistics 2025d)	Open public tables	SSI caseload, recipient composition, payment levels, and state patterns	Aggregate, not household microdata
OASDI Beneficiaries by State and County (Social Security Administration, Office of Research, Evaluation, and Statistics 2025c)	Open public tables	Geographic validation for beneficiary counts and amounts	No full person-level covariates
CPS ASEC (U.S. Census Bureau 2025)	Public microdata	Cross-sectional income, demographic, and Social Security income benchmarks	Retrospective income reporting, limited panel structure
ACS (U.S. Census Bureau 2026)	Public microdata and tabulations	Population geography, demographic structure, and local benchmarking	Limited retirement-program detail relative to SSA files
SIPP (U.S. Census Bureau 2024)	Public longitudinal microdata	Program participation dynamics, family composition, wealth, and monthly transitions	Shorter panel horizon than a lifecycle model
PSID (Panel Study of Income Dynamics 2025)	Public longitudinal microdata	Long-run earnings, wealth, marriage, fertility, and intergenerational structure	Smaller sample, weaker direct program detail
HRS (University of Michigan, Institute for Social Research 2025)	Public longitudinal microdata	Older-age work, claiming, health, and wealth dynamics	Focused on older cohorts
NHATS/NSOC (National Health and Aging Trends Study 2025)	Public or application-based microdata	Functional status, care receipt, and caregiving for LTC extensions	Older-adult focus, not all-age LTSS
MCBS (Centers for Medicare & Medicaid Services 2025a)	Public-use and limited data	Medicare beneficiaries, utilization, health status, and older-age spending	Community/facility split matters for use cases
NVSS mortality tables (Centers for Disease Control and Prevention, National Center for Health Statistics 2025)	Open public tables	Mortality, life expectancy, and cause-of-death structure	Not a direct Social Security mortality table
OACT life tables and Trustees material (Social Security Administration, Office of the Chief Actuary 2025; Board of Trustees, Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds 2025)	Open public tables and reports	Projection alignment, mortality assumptions, taxable payroll, and beneficiary totals	Designed for official projection use, not open micro-simulation replication
T-MSIS overview and TAF documentation (Centers for Medicare & Medicaid Services 2025c, 2025d)	Open documentation plus research files	Medicaid enrollment, LTSS utilization, state variation, and later LTC validation	Data quality varies by state; research files require more setup
MDS 3.0 technical information (Centers for Medicare & Medicaid Services 2025b)	Open technical documentation and files	Institutional care benchmarks for nursing-home populations	Not a household survey; linking to other populations is nontrivial

Minimum validation set by model block

Population and household structure

The baseline population should be judged first against Census products and then against longitudinal surveys.

• Use ACS for geography, age, sex, race, housing, and local population structure (U.S. Census Bureau 2026).
• Use CPS ASEC for household income structure, filing-unit-adjacent relationships, and Social Security income receipt in the base year (U.S. Census Bureau 2025).
• Use SIPP and PSID to validate whether the synthetic population carries plausible family dynamics and wealth heterogeneity, not just plausible cross-sectional margins (U.S. Census Bureau 2024; Panel Study of Income Dynamics 2025).

Earnings, coverage, and insured status

This is where the proposal needs the most discipline. Public validation does not recover exact administrative earnings histories, but it can still test the outputs that matter for Social Security use.

• Use CPS ASEC and SIPP for observed earnings distributions and zero earnings patterns (U.S. Census Bureau 2025, 2024).
• Use PSID for cross-age earnings persistence, volatility, and long-horizon rank movement (Panel Study of Income Dynamics 2025).
• Use the SSA Annual Statistical Supplement to validate insured-worker counts, beneficiary composition, awards, and the downstream results of covered-work histories after the rules are applied (Social Security Administration, Office of Research, Evaluation, and Statistics 2025b).

The proposal should continue to treat public validation of AIME-like and benefit-like outputs as a must-pass requirement even before restricted SSA linkages exist.

Claiming, auxiliary benefits, and beneficiary status

Public SSA tables are stronger here than many teams realize.

• The Annual Statistical Supplement provides national benchmarks for retired workers, spouses, survivors, disabled workers, dual entitlement, beneficiary families, awards, terminations, and representative payees (Social Security Administration, Office of Research, Evaluation, and Statistics 2025b).
• OASDI Beneficiaries by State and County provides a public geographic cross-check once national totals look reasonable (Social Security Administration, Office of Research, Evaluation, and Statistics 2025c).
• CPS ASEC, ACS, SIPP, and PSID help validate the demographic and family side of the history-construction problem, especially marriage, widowhood, age structure, and co-residence (U.S. Census Bureau 2025, 2026, 2024; Panel Study of Income Dynamics 2025).

Disability and SSI

This is one of the strongest public-validation areas because SSA publishes a surprisingly rich disability record.

• Use the DI Annual Statistical Report for awards, application outcomes, diagnosis mix, benefit amounts, return-to-work patterns, and disabled beneficiary structure (Social Security Administration, Office of Research, Evaluation, and Statistics 2025a).
• Use the Annual Statistical Supplement for disability and SSI counts embedded in the broader OASDI/SSI system (Social Security Administration, Office of Research, Evaluation, and Statistics 2025b).
• Use the SSI Annual Statistical Report for recipient composition, payment distributions, and blindness/disability basis categories (Social Security Administration, Office of Research, Evaluation, and Statistics 2025d).
• Use SIPP for joint means-tested program participation and household income context around SSI and disability receipt (U.S. Census Bureau 2024).

Mortality and projection drift

Projection validity should be judged against public official projections, not only against internally generated smooth paths.

• Use the Trustees Report for topline beneficiary, taxable-payroll, and trust-fund alignment targets (Board of Trustees, Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds 2025).
• Use OACT life tables for Social Security-relevant mortality structure (Social Security Administration, Office of the Chief Actuary 2025).
• Use NVSS mortality tables to cross-check mortality by age and sex against broader public-health data (Centers for Disease Control and Prevention, National Center for Health Statistics 2025).
• Use ACS and CPS for denominator populations when validating drift in population structure (U.S. Census Bureau 2026, 2025).

LTC and caregiving extension

The LTC extension would also have a substantial public validation stack, even if it is thinner and more fragmented than the Social Security one.

• Use HRS for older-age health, work, wealth, and retirement behavior (University of Michigan, Institute for Social Research 2025).
• Use NHATS/NSOC for ADLs, IADLs, cognitive impairment, care setting, and family caregiving (National Health and Aging Trends Study 2025).
• Use MCBS for Medicare beneficiaries’ health, utilization, and spending patterns over time (Centers for Medicare & Medicaid Services 2025a).
• Use T-MSIS and TAF for Medicaid LTSS participation and state variation (Centers for Medicare & Medicaid Services 2025c, 2025d).
• Use MDS for nursing-home resident and facility-side functional benchmarks (Centers for Medicare & Medicaid Services 2025b).

This still does not make LTC easy. It does mean that an LTC extension can be validated more seriously than most high-level proposals imply.

Recommended stage-gate use

If this project is funded in phases, the validation stack should be matched to those phases explicitly.

Stage	Public sources that should be sufficient to pass the stage
Stage 0: Base population and synthesis parity	ACS, CPS ASEC, SIPP, PSID
Stage 1: Earnings architecture and benefit baseline	CPS ASEC, SIPP, PSID, SSA Annual Statistical Supplement
Stage 2: Disability, claiming, and auxiliary-benefit realism	SSA Annual Statistical Supplement, DI Annual Statistical Report, SSI Annual Statistical Report, OASDI by State and County
Stage 3: Forward projection credibility	Trustees Report, OACT life tables, NVSS mortality tables, ACS geography checks
Optional LTC pilot	Colorado state rule packet, HRS, NHATS/NSOC, MCBS, T-MSIS/TAF, MDS

That framing matters for funders. It makes clear that the project can be measured against public evidence all along the way.

What public sources still cannot solve

This appendix should not be read as claiming that public validation is enough for every question.

• Public sources will not reproduce exact SSA administrative earnings histories at the person level.
• Public sources are weaker on precise application processing, adjudication timing, and some program-interaction edge cases.
• LTC remains harder because the public record is split across household surveys, facility instruments, Medicaid systems, and state manuals.

But those limitations are different from saying the model cannot be validated. The more accurate statement is that the project can produce a substantial public validation record before it reaches the frontier where restricted administrative data would add the most value.

Board of Trustees, Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds. 2025. The 2025 Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds. Social Security Administration. https://www.ssa.gov/oact/TR/2025/tr2025.pdf.

Centers for Disease Control and Prevention, National Center for Health Statistics. 2025. Mortality Tables. https://www.cdc.gov/nchs/nvss/mortality_tables.htm.

Centers for Medicare & Medicaid Services. 2025a. Medicare Current Beneficiary Survey (MCBS). https://www.cms.gov/data-research/research/medicare-current-beneficiary-survey.

Centers for Medicare & Medicaid Services. 2025b. Minimum Data Set (MDS) 3.0 Technical Information. https://www.cms.gov/medicare/quality/nursing-home-improvement/minimum-data-set-technical-information.

Centers for Medicare & Medicaid Services. 2025c. Transformed Medicaid Statistical Information System (t-MSIS). https://www.medicaid.gov/medicaid/data-systems/medicaid-and-chip-business-information-solution/transformed-medicaid-statistical-information-system-t-msis.

Centers for Medicare & Medicaid Services. 2025d. Transformed Medicaid Statistical Information System (t-MSIS) Analytic Files (TAF). https://www.medicaid.gov/medicaid/data-systems/macbis/medicaid-chip-research-files/transformed-medicaid-statistical-information-system-t-msis-analytic-files-taf.

National Health and Aging Trends Study. 2025. National Health and Aging Trends Study (NHATS). https://www.nhats.org/nhats.

Panel Study of Income Dynamics. 2025. PSID Home. https://psidonline.isr.umich.edu/.

Social Security Administration, Office of Research, Evaluation, and Statistics. 2025a. Annual Statistical Report on the Social Security Disability Insurance Program, 2024. https://www.ssa.gov/policy/docs/statcomps/di_asr/index.html.

Social Security Administration, Office of Research, Evaluation, and Statistics. 2025b. Annual Statistical Supplement, 2025. https://www.ssa.gov/policy/docs/statcomps/supplement/2025/.

Social Security Administration, Office of Research, Evaluation, and Statistics. 2025c. OASDI Beneficiaries by State and County, 2024. https://www.ssa.gov/policy/docs/statcomps/oasdi_sc/.

Social Security Administration, Office of Research, Evaluation, and Statistics. 2025d. SSI Annual Statistical Report, 2024. https://www.ssa.gov/policy/docs/statcomps/ssi_asr/.

Social Security Administration, Office of the Chief Actuary. 2025. Actuarial Life Table, 2022, as Used in the 2025 Trustees Report. https://www.ssa.gov/oact/STATS/table4c6.html.

University of Michigan, Institute for Social Research. 2025. The Health and Retirement Study. https://hrs.isr.umich.edu/.

U.S. Census Bureau. 2024. Survey of Income and Program Participation (SIPP). https://www.census.gov/programs-surveys/sipp.html.

U.S. Census Bureau. 2025. Annual Social and Economic Supplements. https://www.census.gov/data/datasets/time-series/demo/cps/cps-asec.html.

U.S. Census Bureau. 2026. American Community Survey (ACS). https://www.census.gov/programs-surveys/acs.