| County | Tracts | Population | Modelled AAL | Mean denial rate | Compound-burden tracts | Moran’s I (AAL) |
|---|---|---|---|---|---|---|
| Ocean | 145 | 638,691 | $48.6M | 14.8% | 11 | 0.36 |
| Cape May | 33 | 95,456 | $44.4M | 17.0% | 9 | 0.32 |
| Hudson | 183 | 712,029 | $41.8M | 21.2% | 18 | 0.53 |
| Monmouth | 155 | 643,064 | $16.4M | 14.9% | 13 | 0.33 |
| Bergen | 203 | 953,243 | $14.3M | 15.7% | 20 | 0.25 |
| Atlantic | 74 | 274,339 | $9.3M | 22.1% | 7 | 0.45 |
| Middlesex | 192 | 860,147 | $6.4M | 18.8% | 14 | 0.20 |
| Union | 120 | 572,079 | $6.0M | 19.7% | 16 | 0.34 |
| Essex | 211 | 853,374 | $5.1M | 24.4% | 11 | 0.24 |
| Burlington | 117 | 461,853 | $4.6M | 15.2% | 8 | 0.41 |
| Passaic | 120 | 519,986 | $3.8M | 21.5% | 8 | 0.27 |
| Somerset | 74 | 344,978 | $3.5M | 13.5% | 7 | 0.21 |
| Mercer | 84 | 383,732 | $2.8M | 19.0% | 7 | 0.41 |
| Camden | 129 | 522,581 | $1.9M | 17.9% | 8 | 0.08 |
| Morris | 110 | 508,816 | $1.8M | 11.9% | 0 | 0.31 |
| Gloucester | 69 | 302,621 | $1.3M | 15.8% | 4 | 0.32 |
| Salem | 25 | 64,840 | $1.2M | 15.5% | 1 | 0.23 |
| Hunterdon | 30 | 129,099 | $1.1M | 12.6% | 1 | 0.09 |
| Sussex | 42 | 144,808 | $804K | 17.3% | 2 | 0.15 |
| Warren | 23 | 109,739 | $703K | 16.4% | 1 | 0.10 |
| Cumberland | 42 | 153,588 | $623K | 21.7% | 4 | 0.08 |
Modelled AAL: AAL/HU = sfha_frac × 0.01 × median home value (fallback $250k, floor $50k) × 0.10 —
a relative-ranking surface. Validated against 202,156 NJ NFIP claims (1978–2026): tract-level rank
correlation ρ = 0.55 (p < 1e-170) — strong, but below this project’s 0.60 target, so the atlas
ranks flood loss well without being a calibrated dollar estimate. Absolute totals scale ~20× across
plausible loss parameters; rankings do not change at all.
AAL/HU is null for zero-housing tracts. Denial rates are hidden for tracts with
<10 applications; tracts with no HMDA records show as no-data. Compound burden = top tercile of both
AAL and denial rate (co-location, not causation). Whether flood exposure predicts denial at all
depends on analytical choices: across the four combinations of loan universe and exposure measure we
tested, two are significant in opposite directions and two are null. No directional
flood-to-credit claim is supported — see output/M4_FINDINGS.md.
Sources: FEMA NFHL · HMDA 2022 (CFPB) · ACS 2022 5-yr · CDC SVI 2020 · TIGER 2022. Inputs public domain / CC-BY; derived atlas CC-BY-4.0.