Expose federal vs. state budgetary impact in compare.py

Context

PolicyEngine-US now models federal/state cost attribution for Medicaid and CHIP via medicaid_federal_cost / medicaid_state_cost / chip_federal_cost / chip_state_cost, plus cross-program aggregates federal_benefit_cost / state_benefit_cost. See PolicyEngine/policyengine-us#8075 and #8076.

Today, policyengine_api/endpoints/economy/compare.py::budgetary_impact returns a single scalar:

budgetary_impact = tax_revenue_impact - benefit_spending_impact

This conflates federal and state fiscal effects. A reform that shifts cost between federal and state (e.g., Medicaid expansion rollback, OBBBA SNAP state-match, FMAP changes) currently shows as a single number that doesn't answer the policy question "what does this cost the federal government vs. states?"

Proposed change

Extend budgetary_impact to return:

{
  "budgetary_impact": <total>,
  "federal_budgetary_impact": <tax_revenue_change - federal_benefit_spending_change>,
  "state_budgetary_impact": <state_supplement_change - state_benefit_spending_change>,
  ...existing keys...
}

Where:

• federal_benefit_spending_change comes from aggregating federal_benefit_cost across the microsim (baseline vs. reform)
• state_benefit_spending_change from state_benefit_cost
• Tax revenue is overwhelmingly federal (federal income tax, payroll tax); state income taxes attributed to states. If the economy-summary payload already splits tax_revenue by level, use that; otherwise add per-level aggregates to the economy summary.

Why this matters

CBO-style reform scoring, state budget impact analysis, and any comparative advocacy depend on separating federal and state fiscal effects. The per-program microsim work is done in PE-US; this is the API-layer surface to expose it.

Related

• PolicyEngine/policyengine-us#8075 (scoping)
• PolicyEngine/policyengine-us#8076 (Medicaid + CHIP, merged/in-review)
• Follow-up issue for policyengine-app-v2 UI

Read policyengine_api/endpoints/economy/compare.py, starting at budgetary_impact, and inspect the economy-summary payload for any existing tax split. Trace how federal_benefit_cost and state_benefit_cost would be aggregated between baseline and reform. Done means the response preserves existing keys and exposes total, federal, and state budgetary impacts with the requested attribution.