Signal Desert Index
Composite 0–100: loan rate (35%), FDIC bank branches (20%), median income (20%), poverty rate (15%), SBDC distance (10%). Bank branches = all FDIC-insured institutions (SoD 2022), not SBA-specific lenders. Lender Desert layer separately measures distance to PA-based SBA 7(a) lender headquarters from FOIA data — a proximity proxy for SBA-specific access, not a direct measure of whether lending is available. Businesses may access SBA loans through lenders headquartered outside the county.
All counts are SBA
approvals, not funds disbursed. Of 13,199 geocoded records, 1,705 (12.9%) carry status CANCLD (cancelled) and 214 CHGOFF (charged off); 8,321 are EXEMPT, 2,102 paid in full, 857 committed. Approval counts therefore overstate capital actually deployed. County aggregates total 13,205 approvals; 13,199 geocoded to a street address, with 6 excluded for incomplete address data (Bucks 1, Cumberland 1, Lancaster 1, Luzerne 1, Philadelphia 2).
Data SBA FOIA 7(a) 2026-06-30 · Census CBP 2022 · Census ACS 2022 · FDIC SoD 2022 · Census Geocoder (100% match)
Regression model OLS R²=0.982. Predictors: business establishment count (Census CBP), median household income (ACS), FDIC branch density, poverty rate. Establishment count implicitly controls for population scale — per-establishment normalization means the model accounts for county size. Predicted volume = structural expectation; gap = actual minus predicted.
Author Leo Sun · The Shipley School, Philadelphia
leosun582@gmail.com