Small-business loan delinquency roughly doubles to triples after a federally declared disaster, and takes about four years to stabilize — longer for firms that don't get emergency credit

Summary

AskAmerica's warehouse does not carry a small-business loan delinquency series, and its one SBA loan table (fiscal.sba_loan_approvals) covers only the ordinary 7(a) and 504 lending programs, not the SBA's disaster (EIDL) loans, so this cannot be answered as a direct warehouse query. Pulling from the two most directly on-point primary studies instead: at the level of individual firms that apply for an SBA disaster recovery loan, the share of their debt that is severely delinquent (90+ days late) roughly doubles after the disaster for firms that get the loan and roughly triples for firms whose application is declined, relative to the firm's own pre-disaster level. For firms that receive emergency credit, that distress stabilizes by about the fourth year after the disaster; for firms that are declined, delinquency (and the related bankruptcy rate) stays elevated for at least seven years — the full window the study could observe — with no clear sign of full recovery. At the much coarser level of an entire bank's loan book (all borrowers in a FEMA-declared county, not just disaster-hit small businesses), the effect is far smaller and diluted: charge-off rates rise only about 10-13% relative to average, and only with a lag of three to five years, not immediately — not enough to threaten bank solvency.

Why this isn't a warehouse-native answer

search_catalog returned no table or column matching "delinquency" anywhere in the corpus (flagged explicitly as an unmatched term). The closest table, fiscal.sba_loan_approvals, was checked directly: it holds 1.96M 7(a) loans and 229K 504 loans (confirmed by SELECT program, COUNT(*) GROUP BY program), both general-purpose SBA lending programs — it carries no rows from the SBA's actual disaster-loan (EIDL) program, which is the portfolio every study of post-disaster small-business credit distress actually analyzes. Even if it did, the table has a loan_status field (approved / paid in full / charged off) but no time-series of delinquency status over a loan's life, so a recovery-time analysis (when did distress start, when did it end) isn't constructible from it. This is a genuine, structural coverage gap, not a query that failed — the answer below is built entirely from primary research literature instead.

Firm-level evidence: how far delinquency rises, and how long it takes to stabilize

The most directly on-point study is Collier, Howell & Rendell, "After the Storm: How Emergency Liquidity Helps Businesses Following Natural Disasters" (NBER Working Paper 32326, April 2024, revised September 2026). It merges 167,000+ SBA disaster-loan applications (1,900 distinct disasters) with Census Longitudinal Business Database records and Experian credit-bureau reports, tracking firms for seven years after the disaster, and identifies causal effects using discontinuities in the SBA's credit-score approval thresholds (a fuzzy regression-discontinuity design).

• Magnitude of the rise. Comparing raw pre- vs post-disaster means (Figure 6, Panel A): "after the disaster, both approved and rejected applicants experience large increases in the delinquent share of debt... the delinquent share roughly triples for the declined applicants and doubles for the approved applicants." The regression estimate: receiving a recovery loan reduces the 90+ day delinquent share of debt by 33.9 percentage points against a sample mean of 13.4% — "more than twice the mean" — and cuts days-late-on-contracts by 35 days against a mean of 16.7 days (a 200%-of-mean effect).
• Bankruptcy moves the same way and is reported with exact pre/post figures: bankruptcy filings rise "from below 2% before the disaster to over 6% afterward" among declined firms, versus a rise to only about 4% among approved firms — the chart above plots these three points directly from the paper's text.
• Time to recover. In the event-study version of the model, the delinquent-debt-share and days-late effects "decline before stabilizing in the fourth year after the disaster year," with no pre-trend beforehand — i.e., roughly a four-year path back to a new steady state for firms that get emergency credit. For firms that are declined, the raw-means chart shows credit delinquencies "grow among denied firms and remain elevated even seven years after the disaster" — the paper's full observation window — so there is no clear evidence of recovery within seven years for that group.

Bank-level evidence: a much smaller, system-wide effect

Blickle, Hamerling & Morgan, "How Bad Are Weather Disasters for Banks?" (Federal Reserve Bank of New York Staff Report 990, revised October 2025) takes the opposite vantage point: not individual disaster-affected borrowers, but an entire bank's loan book (all customers, disaster-hit or not) in counties that received a FEMA weather-disaster declaration, 1995-2018. This dilutes the effect enormously relative to the firm-level study — which is itself informative about scale.

• For the average FEMA disaster, multi-county bank charge-off rates rise "with a delay" and only reach significance at the five-year horizon, cumulating to about a 10% increase relative to the average charge-off rate; single-county (hyper-local) bank charge-offs and default risk (Z-score) show no significant baseline effect.
• For the most destructive disasters (95th percentile of damages), single-county bank charge-offs rise significantly in years one and three, cumulating to about a 12% increase by year three, and their Z-score (inverse default-risk measure) falls about 9% relative to average by year five.
• The paper's overall verdict: "disaster exposure may increase credit losses, with a delay, by as much as ten percent relative to average... [but] does not appear to threaten bank survival." A future disaster would need to be roughly 6.7 standard deviations more destructive than the worst disasters in the 1995-2018 sample to push an average single-county bank toward failure.

The gap between the two studies' magnitudes (roughly doubling/tripling for disaster-hit applicants vs. roughly 10-13% for a bank's whole book) is itself informative: post-disaster small-business credit distress is real and large for the businesses actually affected, but it is a small slice of any given bank's total lending, so it barely shows up in bank-wide performance data — consistent with the same NBER paper's finding that recovery loans "crowd in" rather than substitute for private credit, and with both papers' emphasis that FEMA/SBA disaster aid materially dampens the shock before it reaches bank balance sheets.

Caveats

• The firm-level numbers are specifically for firms that applied to the SBA's disaster loan program — a self-selected group of businesses damaged enough to seek federal recovery credit — not a random sample of every small business in a declared-disaster county. The true county-wide average rise in delinquency (including businesses that never apply for SBA aid) is very likely smaller than the applicant-level doubling/tripling figure, and probably sits somewhere between the firm-level and bank-level magnitudes reported here, but no source in this corpus or in the literature reviewed directly measures that county-wide average.
• The NBER paper's "doubles/triples" comparison is read off a described raw-means chart (Figure 6, Panel A) rather than a table of exact digitized values, since the source is a PDF chart, not underlying data; the regression coefficients quoted alongside it (33.9pp reduction, 35-day reduction, exact bankruptcy percentages) are exact figures from the paper's tables and text.
• Neither study reports a single national "average recovery time" number; "about four years to stabilize" describes the point at which the treatment-effect estimate flattens in the event-study figure for aided firms, not a guarantee that every disaster-hit county or firm recovers on that exact timeline.

What This Report Does Not Answer

• Grain: 'in counties that get hit' implies a county-wide, not just applicant-level, measure: No source found (in this warehouse or the literature reviewed) measures delinquency for the full population of small businesses in a disaster county — only for the subset that applies for SBA disaster credit (firm-level) or for an entire bank's loan book including undamaged borrowers (bank-level). This is disclosed explicitly in the Caveats section as a gap between the available data and the exact grain the question names.

Every query behind this report

1 warehouse call ran in this session, in order. Each is reproducible against the same snapshot.

SELECT program, type, COUNT(*) AS n FROM fiscal.sba_loan_approvals GROUP BY program, type ORDER BY n DESC LIMIT 20