No dataset — including this corpus — reports small-business loan approval rates by census-tract racial composition; the closest evidence points to roughly 20-50% less lending activity in majority-minority tracts

Summary

No dataset anywhere — including this corpus — currently reports small-business loan applications and denials broken out by census-tract racial composition, so a true "approval rate" gap cannot be computed. Confirmed live: the one federal source with small-business lending at census-tract grain, CRA aggregate data (banking.cra_small_business_lending in this corpus, 2009–2020, 879,084 rows), reports only loans originated — it has no applications-received or denials column, so there is no denominator to build an approval rate from, at any grain. The Consumer Financial Protection Bureau's Section 1071 rule, enacted specifically to collect small-business credit applications with tract and denial data together, is still phasing in compliance (largest lenders since July 2025, mid-size lenders January 2026, smaller lenders October 2026); the CFPB does not expect to review this data until mid-2026 and no bulk tract-level results are public yet.

Using the best available proxies, published studies consistently find majority-minority census tracts receive noticeably less small-business credit than other tracts in the same metro area. The most directly relevant primary data point: an analysis of over 5 million Paycheck Protection Program (PPP) loans by Reveal/Center for Investigative Reporting and the Los Angeles Times, published in the LA Times (fetched directly, 2021), found that in the Los Angeles metro area, businesses in majority-white census tracts received PPP loans at twice the rate of majority-Latino tracts, 1.5 times the rate of majority-Black tracts, and 1.2 times the rate of majority-Asian tracts — all within the same metro. Nationally, the same analysis found that in the vast majority of metro areas with population over 1 million, majority-white tracts had a higher lending rate than any tract with a Latino, Black, or Asian majority. Translated into a "how much smaller" framing, that puts majority-minority tracts at roughly 20% (Asian-majority) to 50% (Latino-majority) lower loans-per-business than majority-white tracts in the same metro in this dataset — but this is loans-per-business, a proxy for credit access, not a formal application-approval rate, since PPP loans (like CRA data) carry no standardized denial record either.

What this corpus can and cannot answer

banking.cra_small_business_lending (confirmed via describe_table) carries loans_lt_100k_count, loans_100k_to_250k_count, loans_250k_to_1m_count, and the small-business-revenue subset — all counts of loans made, by census tract, county, MSA and income tier. There is no applications-received or denials column anywhere in this table, confirmed directly against its schema, so no approval rate can be derived from it no matter how the SQL is written.

This corpus also has no census-tract-level race/ethnicity table: census.acs_race_ethnicity and census.acs_population are state/county grain only. The one place a tract-level minority-population-percent field exists in this corpus is housing.hmda_lending_by_tract — which does carry real application, origination, and denial counts — but that table covers home mortgages for 2024 only, a different credit market and a four-year vintage gap from the CRA small-business table's 2020 endpoint. Joining the two to manufacture a single number would conflate mortgage lending with small-business lending and misalign vintages by four years, so it is reported here as a genuine, confirmed corpus gap rather than bridged into a single figure.

What the published record actually shows instead

• PPP loans by tract racial majority, Los Angeles metro (Reveal/CIR + LA Times, 2021, fetched directly): majority-white tracts received loans at 2x the rate of majority-Latino tracts, 1.5x majority-Black, 1.2x majority-Asian; nationally, majority-white tracts out-lent every minority-majority tract type in most metros over 1 million population. Neighborhood examples cited in the article: Playa del Rey (majority-white) 61% of businesses received a PPP loan vs. Inglewood (Black enclave) roughly half that rate, vs. a predominantly Latino South L.A. tract at only 10%.
• NCRC "Disinvestment, Discouragement and Inequity in Small Business Lending" (2008–2016, 7 metro areas, fetched directly): business owners in wealthier tracts received the largest share of CRA-reported loans — 85% in Milwaukee — and in 6 of 7 metros studied, over 70% of loans went to middle- and upper-income neighborhoods; SBA 7(a) lending to Black-owned businesses fell from ~8% to ~3% of loans during the Great Recession and has not recovered.
• Brookings/Johns Hopkins 21st Century Cities analysis of Black-majority communities (2021, fetched directly): by 2021, majority-Black census tracts were markedly less likely to have a bank branch at all than non-majority-Black tracts (mapped concretely in Philadelphia); majority-Black ZIP codes in large metros had a median banking-competition HHI of 4,584 vs. 3,106 in non-majority-Black ZIP codes (higher HHI = less competition, typically worse loan terms).
• CFPB matched-pair testing pilot (Fairfax County, VA and Nassau County, NY, fetched directly from CFPB's official release): not a tract-composition study, but the closest thing to a controlled experiment on differential treatment — lenders expressed interest in lending to 40% of white test participants but only 23% of Black participants who had equal-or-better financial profiles, a 17-point gap in willingness to lend before a formal application was ever filed.
• Chicago region, 2008–2012 (Woodstock Institute): only ~29% of businesses in low-income majority-minority tracts received a loan vs. ~60% in upper-income majority-minority tracts and ~75% in upper-income majority-white tracts — showing tract income tier compounds but does not fully explain the racial-composition effect.

Bottom line

If forced to give one number: the best tract-level evidence (PPP loans by census-tract racial majority, Los Angeles metro) puts majority-minority tracts at roughly 20% to 50% lower small-business lending activity than majority-white tracts in the same metro, with the size of the gap varying by minority group, city, and the correlated effect of tract income level. This is a lending-intensity proxy (loans per business), not a true application-approval rate, because no dataset currently in existence — public or in this corpus — reports small-business loan applications and denials broken out by census tract. Section 1071 was written specifically to close this gap; its data is not yet available at the scale needed to answer this question directly.

What This Report Does Not Answer

• A quantified 'how much smaller' approval-rate figure: Confirmed via describe_table that no public dataset (including this corpus's CRA small-business table) reports small-business loan applications or denials at census-tract grain, so a true approval rate cannot be computed by anyone currently, including CFPB's own Section 1071 program which is still phasing in. Reported the closest available primary-sourced proxy (PPP loans-per-business by tract, fetched directly from LA Times/Reveal) with a real range (roughly 20-50% lower lending activity in majority-minority tracts) instead of fabricating a precise approval-rate percentage.