Benchmark model component matrix

Why this chapter exists

The existing-models.md chapter is strategic. It answers the question “where does this project fit in the landscape?”

This chapter is more operational. It answers a narrower question:

• how do the main benchmark models appear to construct the pieces we care about?
• which parts are documented publicly, and which are not?
• what should this proposal copy, avoid, or defer?

That distinction matters because the grant case is not just “we think a public model would be useful.” It is “we understand the current benchmarks well enough to know what work the funding would actually pay for.”

Scope and interpretation

This matrix is based on the public record, not on private access or inference about internal codebases.

That means:

• if a capability is well documented publicly, the matrix says so
• if a capability likely exists but the public record is thin, the matrix says the public record is thin
• if a model is adjacent rather than directly comparable, the matrix says that too

This chapter covers five comparison objects:

1. Our plan longitudinal populace plus PolicyEngine-US plus this Social Security application layer
2. DYNASIM the main non-governmental dynamic benchmark
3. MINT the main administrative-data benchmark for Social Security analysis
4. CBO / CBOLT the official long-term fiscal benchmark in public discussion
5. Morningstar an adjacent retirement-adequacy benchmark that matters for wealth, retirement, and LTSS but is not a direct Social Security scoring model

Morningstar is included because Gopi’s benchmark set included it and because its recent public work is one of the clearest examples of a modern retirement-outcomes model being used for LTSS policy analysis (Morningstar 2024; Look and VanDerhei 2025a, 2025b).

High-level orientation

Model	Primary objective	Public access	Main strength	Main limitation for our purposes
Our plan	Build a public longitudinal population asset and a transparent Social Security application layer	Intended to be open-source and publicly documented	Transparency, inspectability, integration with PolicyEngine	Must prove that public synthetic construction is good enough
DYNASIM	Broad retirement-income and aging microsimulation	Documentation and outputs are public; code is not	Breadth, maturity, family and LTSS scope (Favreault et al. 2015; Urban Institute 2024)	Not independently reproducible
MINT	Social Security, SSI, and retirement-income analysis with administrative-data credibility	Public methodology; restricted data and code	Administrative earnings credibility (Smith et al. 2010; Smith et al. 2021; Social Security Administration 2024)	Not a public model and simplified in some behavioral margins
CBO / CBOLT	Official long-term budget and Social Security outlook	Public reports, not public production microdata/code	Institutional authority and macro-fiscal integration (Congressional Budget Office 2004, 2018, 2024b, 2024a)	Public record is relatively thin on record-level construction details
Morningstar	Retirement-income adequacy and retirement-product or policy analysis	Public papers and technical notes; proprietary model	Household retirement adequacy, assets, and recent LTSS work (Look and VanDerhei 2024, 2025a, 2025b; Morningstar 2024)	Not a direct public Social Security microsimulation benchmark

Matrix 1: population and state construction

Component	Our plan	DYNASIM	MINT	CBO / CBOLT	Morningstar
Base population	Public synthetic populace, PolicyEngine’s ML-first microdata layer, extended longitudinally	SIPP-based starting sample with publicly documented 0.04% core and 0.4% expanded variants (Favreault et al. 2015; Urban Institute 2024)	SIPP plus administrative earnings and program records for strong near-retirement credibility (Smith et al. 2010; Smith et al. 2021; Social Security Administration 2024)	SSA Continuous Work History Sample foundation, with a 1-in-1,000 representative microsimulation sample and SIPP/CPS imputations for missing demographics and family structure (Congressional Budget Office 2018, 2019)	Household-oriented retirement simulation using current resources, projected longevity, healthcare, and retirement assets; public materials emphasize model outputs more than raw starting-file mechanics (Look and VanDerhei 2024; Morningstar 2024)
Historical earnings	Synthetic reconstruction from panel data, calibrated against populace’s registry of administrative targets; benchmarked across candidate model families	Public record shows lifetime economic histories built from survey and linked administrative inputs, with annual updating and alignment (Favreault et al. 2015; Urban Institute 2024)	Strongest public benchmark because administrative earnings are built in for many cohorts (Smith et al. 2010; Social Security Administration 2024)	Public record discusses lifetime earnings assumptions and fiscal outputs, but is much less explicit on the micro history-construction machinery (Congressional Budget Office 2004, 2018, 2024b)	Public papers say the model estimates historical wages for each household member and then simulates accumulation and retirement adequacy; claim age is simplified in the inaugural analysis (Look and VanDerhei 2024)
Family structure	Explicit relationship-history layer with spouse links, widowhood, divorce duration, remarriage, and benefit-facing auxiliary states	Public documentation shows marriage, divorce, family structure, and spouse-related states are part of the annual simulation (Favreault et al. 2015; Urban Institute 2024)	Public methodology supports spouse and survivor benefit analysis, but the public record is less explicit than DYNASIM on relationship-history mechanics (Smith et al. 2010; Social Security Administration 2024)	Public record is relatively thin on family-history construction at the record level (Congressional Budget Office 2004, 2018)	Public outputs are household-based and broken out by family status, but the public record does not suggest a fully general spouse-former-spouse-child network like the one needed for detailed auxiliary-benefit analysis (Morningstar 2024; Look and VanDerhei 2024)
Disability and health	Separate impairment, program-pathway, and claiming states, plus mortality and family interactions	Publicly documented health, disability, cognition, and work-limitation modules with yearly transitions (Favreault et al. 2015; Urban Institute 2024)	Includes disability pathways but with publicly documented simplifications around adjudication and return-to-work rules (Social Security Administration 2024)	Public record is strong on aggregate Social Security finances and disability spending, weaker on record-level disability-state machinery (Congressional Budget Office 2024b, 2024a)	Public papers explicitly include healthcare costs, projected longevity, and LTSS states such as home healthcare and nursing home need, but not a public SSDI-style program pathway (Look and VanDerhei 2024, 2025a)
Wealth, assets, and LTSS	Not phase-1 core, but preserved as a later extension track through longitudinal populace	Major documented strength: wealth, pensions, health spending, LTSS use, payer assignment, and Medicaid interaction (Favreault et al. 2015; Urban Institute 2024; Favreault 2020)	Stronger than a Social Security-only model on pensions and SSI interactions, but not positioned publicly as a leading LTSS model (Smith et al. 2010; Social Security Administration 2024)	Public emphasis is fiscal outlook rather than household adequacy, wealth depletion, or LTSS risk pathways (Congressional Budget Office 2024b, 2024a)	Major strength: retirement assets, expenses, projected inadequacy, and recent LTSS and WISH analyses using the same model family (Look and VanDerhei 2024, 2025a, 2025b)

Matrix 2: benefit logic, behavior, and policy use

Component	Our plan	DYNASIM	MINT	CBO / CBOLT	Morningstar
Social Security rule engine	PolicyEngine-US provides an open rule engine, with this repository adding longitudinal inputs and validation	Public documentation says Social Security and SSI are strongly rule-based components (Favreault et al. 2015; Urban Institute 2024)	Public documentation indicates current-law Social Security, SSI, and pension logic with SSA mission alignment (Smith et al. 2010; Social Security Administration 2024)	Public reports provide policy and fiscal results, but not a fully inspectable public rule engine comparable to PolicyEngine-US (Congressional Budget Office 2018, 2024b)	Public materials indicate Social Security benefits are estimated for each household member, but not a public, inspectable Social Security rules stack (Look and VanDerhei 2024)
Claiming behavior	Explicit claim-age buckets and reduced-form claiming hazards are in scope	Public record indicates a serious retirement-and-program timing model, but not a highly transparent public description of every claiming equation (Favreault et al. 2015; Urban Institute 2024)	Public methodology explicitly notes simplification to a single claiming age and omission of sophisticated claiming strategies (Social Security Administration 2024)	Public CBOLT overview describes annual eligibility checks and claiming probabilities with spikes at age 62, Medicare eligibility age, FRA, and age 70 (Congressional Budget Office 2018)	In the inaugural Morningstar analysis, claim age is assumed to equal retirement age, which is analytically tractable but much simpler than the behavior we need for Social Security microsimulation (Look and VanDerhei 2024)
Auxiliary benefits	Spouse, survivor, divorced spouse, and dual-entitlement logic are treated as a dedicated family-history build problem	Public scope clearly includes family structure and Social Security benefit logic, making spouse and survivor analysis part of the model’s natural domain (Favreault et al. 2015; Urban Institute 2024)	Publicly useful for spouse and survivor outputs, but less explicit on the underlying family-history construction in current public docs (Smith et al. 2010; Social Security Administration 2024)	Publicly visible outputs matter here, but the construction details are not comparably open (Congressional Budget Office 2024b)	Public model outputs are household-based, but the public record is not strong enough to treat Morningstar as an auxiliary-benefit benchmark in the MINT or DYNASIM sense (Look and VanDerhei 2024; Morningstar 2024)
Behavioral responses	Reduced-form claiming, work, and transition responses; no phase-1 promise of full structural optimization	Public model supports retirement-income and policy analysis with rich state transitions and cross-domain behavior (Favreault et al. 2015; Urban Institute 2024)	Stronger on administrative earnings credibility than on open behavioral detail; public docs disclose simplifications where relevant (Social Security Administration 2024)	Strongest on fiscal integration and official assumptions, not on public behavioral transparency (Congressional Budget Office 2018, 2024a)	Strong on retirement timing, contribution behavior, asset accumulation and drawdown, and adequacy outcomes; weaker fit for detailed Social Security claimant-pathway modeling (Look and VanDerhei 2024; Morningstar 2024)
Policy use cases	Transparent Social Security reform analysis first, with extension path to SSI, adequacy, and LTC	Broad retirement, Social Security, SSI, wealth, caregiving, and LTSS policy use (Favreault et al. 2015; Favreault 2020; Johnson et al. 2023)	Social Security, SSI, pensions, retirement-income distribution, and beneficiary analysis (Smith et al. 2010; Social Security Administration 2024)	Official long-term Social Security finances and budget effects (Congressional Budget Office 2024b, 2024a)	Retirement adequacy, plan design, savings policy, and now LTSS adequacy and WISH-style backstop analysis (Look and VanDerhei 2024, 2025a, 2025b)

Matrix 3: alignment, validation, and transparency

Component	Our plan	DYNASIM	MINT	CBO / CBOLT	Morningstar
External alignment	Explicit stage-gated calibration and validation against public targets	Publicly documented alignment to SSA OACT targets for major demographic and labor outcomes (Urban Institute 2024)	Administrative-data benchmarking is built into the model’s institutional setting (Social Security Administration 2024)	Deeply integrated with CBO’s official long-term outlook assumptions (Congressional Budget Office 2018, 2024a)	Public materials describe stochastic scenario modeling and household outcome projections, but not a public OACT-style alignment regime for Social Security outputs (Look and VanDerhei 2024; Morningstar 2024)
Validation emphasis	Public validation artifacts are a core deliverable, not a footnote	Long publication trail and broad institutional credibility	Administrative-data credibility and SSA use case are the central strengths	Official scorekeeping credibility at the aggregate level	Public reports and technical appendices are useful, but validation is oriented toward adequacy findings rather than open benchmarking against SSA-style intermediate states
Transparency	Intended full transparency on methods, code, and validation	Documentation public; full codebase not public	Methodology public; data and code restricted	Reports public; production internals not public	Research papers and technical notes public; full model remains proprietary
What we should learn	N/A	Match the seriousness about state richness and alignment	Treat lifecycle validation as the hardest problem	Do not overclaim official-scoring parity	Remember that retirement adequacy and LTSS policy can be highly decision-relevant even when the Social Security micro-paths are simplified

What this means for each workstream

The matrix is only useful if it changes how we describe the build.

1. Earnings and history construction

The lesson is not that there is one canonical benchmark to copy.

• MINT is the benchmark for why historical earnings matter so much.
• DYNASIM is the benchmark for integrating those histories into a broader lifecycle model.
• CBO reminds us that long-run Social Security analysis lives in a larger fiscal conversation.
• Morningstar shows that modern retirement models can be influential even when they are organized around household adequacy rather than a classic Social Security microsimulation stack.

So the funding case should say: the hard work is building credible public histories and validating them, not merely plugging current earnings into a benefit formula.

2. Family, disability, and claiming

This is where the public comparison becomes especially clarifying.

• DYNASIM is the strongest public benchmark for saying these states need to be first-class objects.
• MINT is useful because it shows that a serious model can still disclose simplifications rather than pretending to do everything.
• Morningstar is useful because it shows that some very influential retirement analysis can proceed with much simpler Social Security timing assumptions.

That combination should make the proposal more disciplined, not less: we should fund a real family-disability-claiming layer, but we should also be explicit about what phase 1 simplifies.

3. Wealth, adequacy, and LTSS

This is the clearest place where the benchmark set splits.

• DYNASIM is far ahead on mature LTSS and payer-path modeling.
• Morningstar is increasingly relevant on retirement-adequacy and LTSS-risk analysis.
• Our plan should not claim near-term parity here.

Instead, the proposal should say that Social Security is the first application and that the architecture preserves a path toward adequacy and LTSS work once the longitudinal population is credible.

4. Transparency as a real differentiator

The matrix also clarifies what “open” should and should not mean.

It should not mean:

• we are automatically more credible than DYNASIM, MINT, or CBO
• public data are just as good as administrative records
• openness substitutes for validation

It should mean:

• methods, assumptions, and failures can be inspected publicly
• benchmark comparisons can be reproduced
• outside users can build on the same population platform

That is a real differentiator. It just is not the same thing as saying we already have the best model.

Why this supports the funding ask

This comparison helps make the actual work legible.

The funding is not for a thin wrapper around existing Social Security rules. It is for:

• longitudinal population construction
• history reconstruction
• family and disability state construction
• validation against multiple benchmark traditions
• integration into an open, inspectable application layer

The benchmarks show that these are substantial model-building tasks, not mere documentation tasks.

Why this is a good candidate for interactivity

This is one of the few parts of the proposal where interactivity would be genuinely useful rather than decorative.

An interactive companion could let readers switch between:

• model
• component
• level of public evidence
• implications for our build

That would make diligence faster for funders and collaborators.

But the written matrix should come first. Interactivity should compress understanding of a settled comparison structure, not substitute for doing the comparison.

Bottom Line

The benchmark landscape is not just “DYNASIM is good, MINT is administrative, CBO is official.”

At the component level:

• DYNASIM is the main benchmark for breadth and state richness
• MINT is the main benchmark for the importance of earnings-history credibility
• CBO is the main benchmark for fiscal authority and macro integration
• Morningstar is the main adjacent benchmark for modern retirement adequacy and publicly visible LTSS analysis

That is the right context for this proposal. The project is strongest if it presents itself as the attempt to build the most transparent public stack in this space, while being explicit that credibility has to be earned component by component.

Congressional Budget Office. 2004. Overview of the Congressional Budget Office Long-Term (CBOLT) Policy Simulation Model. Technical Paper 2004-01. Congressional Budget Office. https://www.cbo.gov/publication/15188.

Congressional Budget Office. 2018. An Overview of CBOLT: The Congressional Budget Office Long-Term Model. Congressional Budget Office. https://www.cbo.gov/publication/53667.

Congressional Budget Office. 2019. Social Security Replacement Rates and Other Benefit Measures: An in-Depth Analysis. Congressional Budget Office. https://www.cbo.gov/publication/55038.

Congressional Budget Office. 2024a. CBO’s 2024 Long-Term Projections for Social Security. Congressional Budget Office. https://www.cbo.gov/publication/60392.

Congressional Budget Office. 2024b. Social Security’s Finances. Congressional Budget Office. https://www.cbo.gov/system/files/2024-05/60281-Social-Security-Finances.pdf.

Favreault, Melissa M. 2020. Incorporating Long-Term Services and Supports in Health Care Proposals: Cost and Distributional Considerations. The Urban Institute. https://www.urban.org/research/publication/incorporating-long-term-services-and-supports-health-care-proposals.

Favreault, Melissa M, Karen E Smith, and Richard W Johnson. 2015. The Dynamic Simulation of Income Model (DYNASIM): An Overview. The Urban Institute. https://www.urban.org/research/publication/dynamic-simulation-income-model-dynasim.

Johnson, Richard W., Karen E. Smith, and Barbara A. Butrica. 2023. Lifetime Employment-Related Costs to Women of Providing Family Care. The Urban Institute. https://www.urban.org/sites/default/files/2025-02/Lifetime-caregiving-costs.pdf.

Look, Spencer U., and Jack VanDerhei. 2024. Beyond the Retirement Crisis Headlines: Why Employer-Sponsored Plans Are the Key to Retirement Adequacy for Today’s Workers. Morningstar Center for Retirement & Policy Studies. https://www.morningstar.com/content/cs-assets/v3/assets/blt9415ea4cc4157833/bltd4bb26598046aed4/66a1535de91a178e5c15872a/Introducing_the_Morningstar_Model_of_US_Retirement_Outcomes_-_July_2024_-_final.pdf.

Look, Spencer U., and Jack VanDerhei. 2025a. The Overlooked Cost: How Long-Term Services and Supports Impacts Retirement-Income Adequacy. https://www.morningstar.com/business/insights/research/ltss.

Look, Spencer U., and Jack VanDerhei. 2025b. WISH Granted: How a National Long-Term Services and Supports Insurance Program Could Boost Retirement Outcomes. https://www.morningstar.com/business/insights/research/wish-act-national-ltss-insurance-program.

Morningstar. 2024. Morningstar Model of u.s. Retirement Outcomes. https://www.morningstar.com/business/lp/model-of-us-retirement-outcomes.

Smith, Karen E, Melissa M Favreault, Barbara A Butrica, and Philip Issa. 2010. Modeling Income in the Near Term 6 (MINT6): Projecting Retirement Income Through 2060. The Urban Institute. https://www.urban.org/research/publication/modeling-income-near-term-mint6.

Smith, Karen E., Aaron R. Williams, and Stipica Mudrazija. 2021. Modeling Income in the Near Term: Version 8. Urban Institute. https://www.urban.org/research/publication/modeling-income-near-term.

Social Security Administration. 2024. Projection Methodology: Modeling Income in the Near Term, Version 8 (MINT8). https://www.ssa.gov/policy/docs/projections/methodology.html.

Urban Institute. 2024. Urban’s Dynamic Simulation of Income Model 4 (DYNASIM4). https://www.urban.org/research/publication/urbans-dynamic-simulation-income-model-4.