Development roadmap
Overview
This roadmap describes a serious, stage-gated build. The goal is not to force a public launch on an arbitrary schedule. The goal is to earn the right to proceed from one stage to the next by passing scoring gates — pre-specified thresholds on the record defined in scoring-and-resolution.md.
The roadmap is therefore organized around capability milestones and decision points, not chronological phases.
Workstreams
The project has four core workstreams:
- Longitudinal Populace construction
- Social Security integration and validation
- Policy analysis products
- Public API and interface
These workstreams overlap, but they should not advance at the same pace. Interface work should lag validation work. Policy analysis should lag baseline replication. That ordering is intentional.
Stage 0: project setup and baselines
Purpose: lock the main project decisions before large-scale implementation.
Deliverables
- Finalized base population platform:
populace, PolicyEngine’s microdata stack - Clear boundary between
populaceplatform work and Social Security-specific application work - Benchmark datasets and target tables assembled from
populace’s primary-source data and its versioned target registry - Initial validation harness for baseline distributions
- Implementation team and external review capacity identified
- Written stage-gate criteria approved internally
- Initial design-partner set and first pilot-analysis memo
Exit criteria
- Clear specification of which outcomes count as success in stage 1
- No unresolved ambiguity about base population or top-level scope
- At least a small set of external users or partner categories identified for the first validated outputs
Stage 1: historical earnings reconstruction
Purpose: determine whether populace can be extended into a credible longitudinal population asset for Social Security analysis.
Core tasks
- Harmonize PSID and related longitudinal sources
- Build at least one conservative production path for earnings-history reconstruction inside
populace - Add the first longitudinal state variables and transition machinery to
populace - Compare alternative model families where justified
- Validate age-earnings profiles, percentiles, mobility, AIME, and correlation structure
- Document where performance is strong and weak
Deliverables
- Longitudinal
populacealpha with earnings histories and core longitudinal states - Validation report on held-out data and external benchmarks
- Recommendation on the production longitudinal architecture
Exit criteria
- Longitudinal
populaceis accurate enough to justify downstream benefit modeling - Validation results are publishable and not merely anecdotal
If these conditions are not met, the project should pause rather than proceeding mechanically.
Stage 2: family, disability, claiming, and benefits
Purpose: turn longitudinal populace into a credible Social Security analysis dataset.
Core tasks
- Freeze the minimal production version of longitudinal
populacechosen at the end of stage 1 - Implement family structure and marital histories needed for auxiliary benefits
- Add disability and mortality transitions
- Specify a minimal but concrete claiming model
- Run synthetic records through PolicyEngine-US benefit logic
- Validate beneficiary counts, average benefits, and key distributions
Deliverables
- Benefit-ready synthetic panel
- Validation report for beneficiary types and benefit amounts
- Replication of a small set of standard baseline tables
- Pilot baseline or reform analyses with clear external use cases
Exit criteria
- Benefit results are close enough to published benchmarks to support exploratory reform analysis
- Known weaknesses are documented and bounded
Stage 3: forward projection and reform analysis
Purpose: move from longitudinal populace plus a validated Social Security layer to a projected dynamic model that can analyze reform packages.
Core tasks
- Add forward aging and cohort entry
- Align near-term and long-term projections to published assumptions
- Test drift control and dynamic calibration
- Replicate selected published reform analyses
- Quantify uncertainty and sensitivity
The operational detail behind this stage now lives in operationalizing-mortality-and-projection-drift.md.
Deliverables
- Forward projection pipeline
- Baseline replication against published projection targets
- Reform-analysis workflows for a defined set of policy packages
- External pilot use of validated baseline or reform workflows
Exit criteria
- Baseline projection quality is strong enough that reform outputs are interpretable
- The model can explain its own uncertainty rather than hiding it
- At least limited outside testing shows the validated outputs are legible and useful to non-team users
Stage 4: public productization
Purpose: expose validated capabilities to outside users.
Core tasks
- Build API endpoints for baseline and reform analysis
- Build a focused public interface
- Publish tutorials, methods notes, and validation artifacts
- Define product boundaries clearly so users know what is and is not reliable
Deliverables
- Public API
- Public web interface
- User documentation
- Public release materials and examples
Exit criteria
- External users can reproduce headline examples
- The interface is narrower than the full model, but more reliable
- Validation artifacts are published alongside the product
Cross-cutting deliverables
Throughout the project:
- maintain versioned validation reports
- preserve reproducible data-processing pipelines where licensing permits
- document model decisions and reversals
- preserve the separation between reusable
populaceinfrastructure and Social Security-specific application code - collect external review from domain experts
Adjacent extension track
The core project is Social Security-first. However, the architecture should preserve a path to adjacent modules that can share the synthetic panel, especially:
- SSI-rich adequacy analysis
- retirement adequacy with wealth and pensions
- long-term care and caregiving policy
The most plausible first LTC step is a static, state-specific rules pilot rather than immediate national dynamic reform scoring. That kind of pilot can prove out the policy-rules layer, scenario outputs, and user demand while the harder national dynamic LTC build remains a later phase.
Those extensions should follow, not precede, validation of the Social Security core.
Summary
The roadmap is deliberately conservative because the main risk in this project is not coding speed. It is false confidence. A public dynamic Social Security model becomes valuable by surviving explicit validation gates, not by reaching a web launch quickly.