Clarify population comparison years and validate sensitivity of headline estimates

Population impact currently shows:

• 16,853,856 qualifying children (31.4% of under-13s).
• Approximately $151.5 billion potential annual subsidy.
• 1,606,000 children served and 9.5 served per 100 who would qualify.
• Qualifying shares of 73.5% in Vermont and 60.6% in Maine.

These outliers are validation targets, not proven errors. The model uses 2026 rules and assumed care, while the observed caseload is an average month of FY2023. The dates appear in explanatory text, but the ratio's own label does not make the cross-year comparison apparent and may be read as a current take-up rate.

Suggested improvements:

• Label the ratio directly as FY2023 served / modeled 2026 qualifying, or align years where feasible; explain it is not a measured current take-up rate.
• Round modeled headline counts to defensible precision (e.g. 16.9 million); retain detailed values in downloadable data if useful.
• Validate high-share states and document sensitivity to care assumptions, activity rules, dataset/state calibration, and counting definitions.
• Keep the hypothetical nature of the $151.5B scenario prominent and document the scope of the “upper bound” claim.

Sources: src/components/Impact.tsx, public/data/impact.json, public/data/acf_served.json.

Start in src/components/Impact.tsx and inspect public/data/impact.json alongside public/data/acf_served.json to trace the displayed years, ratios, counts, and state shares. Compare the modeled 2026 values with the FY2023 served data, then define the validation and sensitivity checks; done means the ratio and hypothetical upper-bound scope are explicit, headline precision is defensible, and high-share states and assumptions are documented.