Allow US calculations in 2036

Goal

Let the US engine and app calculate 2036. CBO's February 2026 baseline runs through 2036, and its 10-year budget window is now 2027–2036. Today the API and app stop at 2035. This is US only; the UK stays at 2024–2030.

Why this is more than the 2035 change

When 2035 was added (#2315, April 2025), it took two lines in policyengine_api/country.py. Today the binding limit is the canonical SPM forecast. It covers 2022–2035, and every layer reads its years from it:

spm-calculator, rolling_forecast_2026_09_09.json;
policyengine-us, through spm.py;
the policyengine.py bundle, which pins the forecast's content_sha256;
the simulation worker's year boundary.

A 2036 request fails with SPM_YEAR_UNAVAILABLE. Everything else already handles 2036:

policyengine-us parameters run to 2100;
the US data release is one 2024-base dataset, uprated per year;
budget windows allow end years up to 2099.

Steps, in release order

Phase 1 (independent):

spm-calculator: extend the forecast horizon to 2036 (build scripts, inputs and artifacts), then release 1.1.0. CBO's 2036 prices are already in cbo_calendar_2026_02.csv.
policyengine-us: load CBO's February 2026 projections for 2036 into the CPI-W, SSA, HHS and SNAP upraters and into the payroll cap and wage index. Move the last-projected-year settings to 2036.
policyengine-api: add 2036 to the US time_period list and the v2 catalog year range; update the docs and the tests that use 2036 as the unavailable year. Merge only together with the bundle bump below.
policyengine-app-v2: move the metadata fallback maxDate to 2036. Everything else reads economy_options.time_period.

Phase 2:

policyengine-us: widen the spm-calculator pin to 1.1.x and update the forecast-years test.
policyengine.py: certify a bundle with spm-calculator 1.1.0, the new forecast hash and a policyengine-us release that includes the pin. Update the docs and tests: 2022–2036 supported, 2037 the unavailable example.

Phase 3 (rollout):

policyengine-api: the bundle bump plus the 2036 PR, in one release.
policyengine-sim-api: the bundle bump, and move the unavailable-year test examples from 2036 to 2037. Deploy the worker before or with the API.

Phase 4 (follow-up):

microcosm: set the annual static aging end_year to 2036, for when policyengine.py Update PolicyEngine US to 0.345.11 #524 activates annual US datasets.

Release-order risk

Shipping the country.py change before a bundle with the 2036 forecast would show 2036 in the year picker while every 2036 run fails. The API PR stays a draft until the bundle PR is ready.

Verification on staging

GET /us/metadata lists 2036.
A national 2036 economy run completes, and its SPM receipt includes "2036".
budget_window:2027:10 completes all ten years.
2037 still returns SPM_YEAR_UNAVAILABLE.
In the app, the year picker shows 2036 and a 2027–2036 window renders.

PRs and issues will be linked here as they open.

🤖 Generated with Claude Code

@MaxGhenis

Linked so far:

policyengine-api: Offer 2036 in the US economy options and v2 catalog #3927 (sub-issue) and Offer 2036 as a US year in economy options and the v2 catalog #3928 (draft PR). Offer 2036 as a US year in economy options and the v2 catalog #3928 stays a draft until a bundle carries the 2036 SPM forecast.
policyengine-app-v2: Fall back to a 2022–2036 date range before metadata loads policyengine-app-v2#1266 moves the metadata fallback maxDate to 2036-12-31. The fallback applies only until metadata loads, so it can merge any time.

Next: the spm-calculator horizon PR. Its CE and ACS components need a rebuild from the raw microdata caches retained from the 9/9 build, so that PR will ask @MaxGhenis to run the two component builds.

spm-calculator (critical path):

Extend the canonical forecast horizon to 2036 spm-calculator#51 (draft) extends the forecast horizon to 2022–2036 and sets the version to 1.1.0.
- It adds scripts/verify_horizon_extension.py, which requires every 2022–2035 value to stay byte-identical and runs the full household grid against the current artifact.
- Widening the horizon changes hashed scientific code, so the CE and ACS components must be rebuilt from the 9/9 raw caches. The PR lists the four commands.
Release spm-calculator 1.1.0 with the 2022–2036 forecast spm-calculator#52 covers the 1.1.0 release. Merging Bump policyengine-us to 0.195.0 #51 publishes it.

Next: the policyengine-us CBO February 2026 refresh, stacked on PolicyEngine/policyengine-us#10225.

@MaxGhenis

Later phases, filed now so the owners can see them:

policyengine.py:
- Certify a bundle with spm-calculator 1.1.0 and the 2022–2036 SPM forecast policyengine.py#618 asks @MaxGhenis and @anth-volk to certify a bundle with spm-calculator 1.1.0, the new forecast digest and a policyengine-us release with the 1.1.x pin. It also asks whether populace_us_2024 needs recertifying, since the bundle certifies policyengine-us 2.2.1.
- Document and test the 2022–2036 SPM forecast horizon policyengine.py#619 (draft) updates the docs and SPM household tests: 2036 is supported and 2037 is the out-of-horizon example. It should merge with the bundle update.
policyengine-sim-api: Move the worker to the bundle with the 2022–2036 SPM forecast policyengine-sim-api#805 asks @anth-volk to bump the worker to that bundle and move the unavailable-year test examples to 2037. The worker deploys before or with the API's bundle bump and Offer 2036 as a US year in economy options and the v2 catalog #3928.
microcosm: Extend US annual static aging to 2036 microcosm#1280 is a non-blocking follow-up to move the annual static aging end_year to 2036, for when policyengine.py#524 activates annual US datasets.

Still to come:

The policyengine-us CBO February 2026 refresh PR, which is in local checks.
The policyengine-us spm-calculator>=1.1.0,<1.2 pin PR, once 1.1.0 is on PyPI.

policyengine-us CBO refresh: PolicyEngine/policyengine-us#10227 (draft) is stacked on PolicyEngine/policyengine-us#10225; I'll retarget it to main after #10225 merges.

It loads CBO's February 2026 projections through 2036 into the SSA, SNAP and HHS upraters, CPI-W, NAWI and the payroll cap. Each series grows from its latest actual at CBO's projected rates.
HHS uprating keeps its 2026 value, so no partner test changes.
It also stops the CPI-U and chained CPI-U extension from overwriting CBO's 2035 point, which applies through 2036.
The PR body has a before/after table for 2027–2037.
PE-US CI runs only on main-based, non-draft PRs. I'm checking the future-year YAML churn locally and will post the results on the PR.

Navigation Menu

Repository navigation

Allow US calculations in 2036 #3926

Goal

Why this is more than the 2035 change

Steps, in release order

Release-order risk

Verification on staging

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Allow US calculations in 2036 #3926

Description

Goal

Why this is more than the 2035 change

Steps, in release order

Release-order risk

Verification on staging

Activity

DTrim99 commented on Oct 14, 2026

DTrim99 commented on Oct 14, 2026

DTrim99 commented on Oct 14, 2026

DTrim99 commented on Oct 14, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions