Our 2025 poverty prediction, graded

[MaxGhenis]

What this adds

Our 2025 poverty prediction, graded: the September 11 SPM poverty registration graded against P60-290, with the diagnosis of the child and senior misses. One article, its chart, a cover, and one posts.json entry (dated 2026-09-17, the grade date). Base is main; nothing else in the diff.

Relationship to other PRs

• Supersedes the prediction post in Add the spm-calculator 1.0 announcement and the 2025 SPM poverty prediction posts #1202, which stays untouched for its spm-calculator announcement. Differences from that branch's version: the cover moves to app/public/assets/posts/, where the blog serves it (the src/data/posts/images copy 404s); the results section uses the direct counterfactual runs (model change = threshold effect + resource effect: all +0.22 = +0.70 − 0.49; children +0.86 = +1.40 − 0.54; 65+ −0.52 = +0.27 − 0.79) instead of inferring the child threshold effect from the gap between the old and new models; it states that all three registered changes reproduce on policyengine-us 2.2.1 (the branch said two); it adds Census's significance markers and margins, the senior-deduction counterfactual and the raw-ASEC aging check; and it corrects the resources paragraph, which listed Social Security, unemployment compensation and energy assistance as computed from statute (in policyengine-us 2.2.1 those are survey inputs; spm_unit_benefits.py).
• Links the companion recap post at /us/research/2025-spm-poverty-at-2024-thresholds (Add the 2025 SPM poverty recap post #1208), so Add the 2025 SPM poverty recap post #1208 should publish first or together.
• Links four files on master of spm-threshold-paper that arrive with PolicyEngine/spm-threshold-paper#10: the grade record, the analysis directory and three results files. Those links 404 until Integrate a testing framework into template site #10 merges. Merge order: spm-threshold-paper#10, then Add the 2025 SPM poverty recap post #1208, then this.
• Does not link the spm-calculator announcement post (still only in Add the spm-calculator 1.0 announcement and the 2025 SPM poverty prediction posts #1202); it links the package repository instead.

Cover

The cover is the graded chart on a 1800×840 canvas, so the card renders. The blog's covers are photographs; a photographic replacement can drop into app/public/assets/posts/2025-spm-poverty-prediction/cover.webp without touching anything else.

Checks

• Mechanism claims audited against the installed policyengine-us 2.2.1 (spm_unit_benefits, spm_unit_capped_housing_subsidy, spm_unit_spm_expenses, spm_unit_health_insurance_premiums, input-variable uprating) and the population release's calibration_diagnostics.json (57,240 households, 5,659 targets, none a poverty rate).
• Every external link resolved (200) except bls.gov, which refuses curl and was checked in a browser; the four spm-threshold-paper links resolve on the PR branch.
• prettier --check passes on the article and posts.json.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
policyengine-calculator-next	Ready	Preview	Sep 29, 2026 11:56am UTC
policyengine-website	Ready	Preview	Sep 29, 2026 11:56am UTC

[MaxGhenis]

Merge audit (2026-09-29): Max ruled to publish this post (d575). Content reviewed in 3 independent Opus rounds (final PASS at 05d3e3a/477ee987). The only change since then is 9fd9e88, moving the post date to its publication day (posts.json: 2026-09-28 14:30 -> 2026-09-29 08:00). gh pr checks exits 0 (12/12) at 9fd9e88, MERGEABLE, not draft, no CHANGES_REQUESTED. The only block is the required approving review, which Max cannot give on his own PR, so this is an admin merge pinned to the reviewed head. spm-threshold-paper #10 merged first, so footnote 7 resolves.