Add --target-family-loss-multiplier to the US fiscal refresh builder#408
Merged
Conversation
The buildi-sparse release (populace-us-2024-buildi-sparse-rmloss100-6e8e929) undershoots 10 SNAP state targets by 7-43% while a per-state feasibility audit shows every one of them is reachable within the 5x weight cap (DC has 1.6x ceiling headroom, IL 2.6x; only CA is marginal at 1.11x). The misses are loss trade-offs, not support capacity: usda_snap holds ~0.9% of the total loss weight under sqrt-value weighting, so small-dollar state targets lose every trade against national income aggregates. This adds a repeatable FAMILY=MULTIPLIER flag that scales the compiled loss weights of one target family and renormalizes the vector to mean 1, so a boosted-SNAP experiment (e.g. usda_snap=8) is a one-flag rebuild. The multipliers are validated at parse time, fail the build when the family matches no compiled target, and are recorded in calibration diagnostics. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Adds a repeatable
--target-family-loss-multiplier FAMILY=MULTflag totools/build_us_fiscal_refresh_release.py, recorded in the build manifest and calibration diagnostics.Why
The buildi-sparse release undershoots 10 SNAP state benefit targets by 7–43% (DC −43%, IL −35%, NM −32%, …; national −8.4%) while a per-state ceiling audit shows every miss is reachable within the 5× weight cap except CA (ceiling/target 1.11). The misses are loss trade-offs, not support capacity:
usda_snapholds ~0.9% of total loss weight under sqrt-value weighting, so small-dollar state targets lose to large national families. A per-family loss multiplier lets a build deliberately re-weight an administratively-anchored family (e.g.usda_snap=8) without touching the global weighting scheme.Cherry-picked onto current main from the audit branch (original commit eca06ea, 2026-07-09); the flag test passes against main's builder.
Held in reserve pending the Build K validation run: the FNS caseload targets (#371/#376) changed the SNAP loss surface, so the multiplier should only be applied if state dollar misses persist after that run.
Refs #370, #372; audit context in the 2026-07-09 SNAP feasibility session.
🤖 Generated with Claude Code