During the Paycheck Protection Program, minority-owned businesses disproportionately borrowed through a nationwide platform channel that forgave loans at markedly lower rates than branch-based community banks. Most of those borrowers were not remote. They sat inside the everyday footprint of a community bank that, on its own record, forgave at a far higher rate.
The zone is every neighborhood within five miles of a community-bank branch, using branch locations from the FDIC Summary of Deposits as they stood in June 2019, before the program began. Loan records come from the SBA's public release. Borrower race is inferred from name and neighborhood, not self-reported — which is the only way to see this at all, since roughly seven in ten borrowers disclosed no race. Inference error makes gaps look smaller than they are, not larger.
The comparison is composition-adjusted. Each neighborhood's expected forgiveness rate is built from its own borrowers' state, industry, loan round and firm size, then raised by the margin community banks nationally achieved above that same expectation. So a neighborhood is never being judged against a different kind of business — only against what community banks did for businesses like its own.
What this is not. It is an accounting comparison, not a causal claim. Borrowers were not assigned to lenders at random, and nothing here establishes that any particular business would have been approved or forgiven had it walked into a branch.
Why sparse areas are pooled. No institution is named or scored here, and every counterfactual uses one national rate rather than any bank's own. But naming no one is not by itself enough: if only one community bank operates near an area, a figure attached to that area describes that bank as surely as a label would. So an area appears only if it clears two tests — at least 50 businesses, and at least three distinct community banks within the same five miles. The second test is the one that matters. A neighborhood can hold hundreds of borrowers and still sit beside a single bank, so counting businesses alone would leave those areas exposed; requiring three institutions removes that possibility by construction. Areas failing either test are pooled — into their county, then their metro area, then their state — until both are met, and the pooled figure is what appears. Pooled areas are drawn as open rings rather than solid marks, and the detail panel says how many neighborhoods went into each. Both tests govern the map and the tooltip identically, so nothing is legible in the shading that is withheld from the text, and they hold at every zoom level.
Why it matters anyway. Distance was not the obstacle — the typical affected borrower sat about two miles from a community-bank branch, and that holds in every kind of neighborhood. What differs is density and throughput: the highest-minority neighborhoods have roughly a third as many community-bank branches within half a mile, and carry about three times as many small businesses per nearby branch. The constraint looks like outreach and capacity at branches that already exist.