What Counts as Small-Business Lending Now
The 2026 Section 1071 rule does not shrink its dataset through one switch. It uses several gates to decide which applications remain visible: the kind of product, the size of the transaction, the lender’s volume, the lender’s regulatory category, and the applicant business’s revenue. This chapter turns those gates into a practical model and then tests the CFPB’s claim that most depository loan activity will still be represented even though far fewer institutions will report.
That distinction matters because “coverage” has more than one denominator. A rule can exclude many small lenders while still capturing most loans by count if larger institutions originate most of them. The same rule can capture less of the dollar volume, fewer rural branches, or entire specialized markets. Understanding which denominator is being used is essential to understanding what “less” means.
Five gates reshape the initial dataset
The CFPB describes its starting point as conventional, generally applicable small-business lending: loans, lines of credit, and credit cards that are likely to be foundational to a business’s formation and operation. The definition of a covered credit transaction still begins broadly—business credit that is not otherwise excluded—but the 2026 rule adds or relies on boundaries that remove several categories from the initial collection.
First gate: product and transaction type. Merchant cash advances (MCAs), agricultural lending, and transactions of $1,000 or less are now excluded. An MCA is described in the rule as a lump-sum payment exchanged for the right to receive a percentage of a small business’s future sales or income up to a ceiling. The Bureau says MCAs can differ from conventional loans in structure, repayment, underwriting, and legal treatment, and that the market is evolving under state-level sales-based-financing rules. It also says that some providers may lack the infrastructure needed for a new reporting regime. On that account, forcing all MCAs into the first dataset could produce information that is difficult to compare or easy to misread.
The exclusion is not a finding that MCAs do not matter. Commenters argued that alternative lenders serve businesses unable to qualify for conventional credit, that MCAs may be high-cost, and that their growth makes them important to fair-lending oversight. The Bureau acknowledged those concerns and said it would monitor the market. Its narrower decision is that it does not yet have a clear enough basis for treating all MCAs alike or for deciding which of them meet ECOA’s definition of credit. The intended sequence is observation first, possible inclusion later.
Agricultural lending is excluded for a related but distinct reason. The rule covers financing for producing crops, fruits, vegetables, and livestock, as well as purchasing or refinancing assets such as farmland, machinery, equipment, breeder livestock, and farm improvements. The CFPB says those loans often depend on biological collateral exposed to weather and disease, seasonal repayment cycles, and underwriting judgments that are difficult to compare with ordinary commercial lending. It also points to overlapping data and oversight through the Farm Credit Administration, the Community Reinvestment Act, and other agricultural programs.
Here too, the counterargument is substantial. Farms are small businesses, community banks are important agricultural lenders, and commenters argued that public data on farm applications and demographic disparities are incomplete. The Bureau’s position is not that existing systems make a full picture available. It is that imposing a new, overlapping regime on specialized and often rural lenders at launch could divert resources from lending and lower the quality of the data that is actually submitted.
The third product boundary is a de minimis threshold: a transaction of $1,000 or less is left out. The CFPB says such transactions are often incidental to a business deposit relationship, including overdraft facilities, and that fixed reporting costs can be disproportionate to the information yielded. The threshold does not remove all microcredit. Transactions from $1,001 upward remain eligible, and the Bureau rejected higher thresholds such as $2,500, $5,000, $25,000, or $50,000 because those would hide financing important to startups, non-employer firms, and the smallest minority-owned businesses. The chosen line is therefore an attempt to filter out ancillary credit without abandoning small-dollar commercial lending altogether.
Second gate: lender volume. The 2023 rule treated a financial institution as covered when it had made at least 100 covered credit transactions to small businesses in each of two preceding calendar years. The 2026 rule raises that threshold to 1,000. This is the most direct way the CFPB concentrates the initial collection in larger, higher-volume lenders. The Bureau says these institutions are more likely to have standardized systems, more resources for compliance, and enough activity to produce statistically useful information while the reporting regime is still being learned.
The rule does not replace the activity threshold with an asset threshold, such as a bank’s total balance sheet. The CFPB’s reason is that lending activity is more directly connected to a lender’s role in the small-business market, while an asset measure would apply unevenly to banks and nondepository lenders. The two-year look-back is meant to give institutions advance notice of possible coverage rather than making a single volatile year decisive.
There is a sharp boundary effect at any numerical threshold. A lender just below 1,000 originations can face no Section 1071 reporting duty while a lender just above it can face the full regime. The CFPB acknowledges that no threshold eliminates this discontinuity. It nevertheless rejects lower thresholds because they would bring many smaller institutions into the system without a proportionate increase in the number of loans represented, and rejects much higher thresholds because they would concentrate reporting too heavily in the largest institutions.
Third gate: lender category. Farm Credit System lenders are excluded even if they meet the volume threshold. The Bureau describes the Farm Credit System as a nationwide network of borrower-owned cooperatives with an agricultural mission, a distinct regulatory environment, and existing reporting obligations. Patronage dividends and limits on eligible borrowers can also make their loan pricing and operations difficult to compare with shareholder-owned commercial banks.
This exemption is contested. Banking and community groups argued that functionally similar agricultural lending should be treated alike and that excluding FCS lenders could distort the picture of credit in rural areas. The CFPB answered that the specialized structure and overlapping Farm Credit Administration reporting make a separate Section 1071 obligation potentially redundant and complex. It says it will continue to monitor the market. Credit unions and community banks, by contrast, remain eligible for coverage; the Bureau says their concerns are addressed mainly by the higher activity threshold rather than by categorical exemptions.
Fourth gate: applicant revenue. The rule reduces the annual gross-revenue ceiling for a covered small business from $5 million to $1 million. The CFPB says a single revenue standard is simple to apply and more likely to focus the first collection on businesses that are truly small and more likely to face credit barriers. It also says businesses under $1 million may include a higher share of minority-owned and women-owned firms than businesses in the $1 million-to-$5 million range.
That is a proxy, not a complete definition of vulnerability. A business with $1.2 million in revenue may still be early in its growth, dependent on credit, or operating in a market where financing is hard to obtain. Commenters argued that lowering the threshold could exclude precisely the firms moving from startup to expansion and could reduce reported lending sharply. The Bureau disputed an estimate that the change would cut reported loans by about half, but the supplied record does not provide a measured post-rule distribution of affected businesses. The revenue line simplifies eligibility at the cost of excluding some growth-stage experiences.
These gates can coexist, and they can exclude an application for independent reasons. Imagine a conventional credit-card application from a business with $800,000 in annual revenue. If the lender has fewer than 1,000 qualifying originations in either look-back year, the application does not enter the Section 1071 dataset even though the applicant satisfies the revenue test. A high-volume lender’s application to a business with $1.4 million in revenue can fail the applicant gate. An FCS lender can remain outside the regime even if its activity would otherwise meet the volume test. An MCA or agricultural loan can fail at the product gate before lender volume or applicant revenue is considered.
The inverse is also important. Indirect lending is not categorically excluded. The CFPB concluded that vehicle, equipment, and private-label credit can help reveal how small businesses obtain essential assets or working capital. Instead of removing those products, the 2026 rule gives lenders more flexibility to collect required demographic information after a credit decision in some situations where an intermediary was the applicant’s first point of contact. The product remains visible; the collection path changes.
The diagram shows why the 2026 change is better understood as a boundary system than as a single reduction in a data form. The CFPB has narrowed the universe before it narrows the information collected about each application. That ordering is central to its incremental strategy: begin with a more manageable population, then decide from experience whether excluded products, lenders, or fields should be brought back.
Why “most loans” can still mean less visibility
The CFPB estimates that the higher threshold will leave roughly 172 to 181 depository institutions subject to reporting. That is only about 1.85% to 1.95% of all depository institutions in the agency’s estimate, including approximately 167 to 176 banks and savings associations and five credit unions. Yet the Bureau projects that these institutions will account for about 91.9% to 92.8% of the number of small-business loan originations by depository institutions.
The apparent paradox comes from concentration. The institutions that make the most loans can account for most loan applications even when they are a small minority of all institutions. Under the Bureau’s model, between 1,421 and 1,570 depository institutions that would have been covered by the 2023 baseline will no longer report. Those institutions represent only about 5.0% to 5.7% of the number of depository small-business loan originations, according to the estimate. On a dollar basis, however, the excluded institutions represent roughly 24.1% to 26.1% of the volume.
That difference changes the question a dataset can answer. If the question is, “What happens in the largest share of applications by count?” the CFPB’s estimate suggests the narrower regime may retain substantial reach. If the question is, “What financing do smaller relationship lenders provide, and how much capital does that represent?” the loss is much larger. Loan count, dollar value, institution count, and applicant or product diversity are not interchangeable measures of representativeness.
The same issue appears in nondepository lending. The Bureau says its information is not comprehensive enough to estimate coverage with the same precision. In its model, about 610 nondepository institutions would have been covered by the 2023 rule, including online lenders, community development financial institutions, merchant cash advance providers, commercial finance companies, governmental lending entities, and Farm Credit System members. The 2026 framework assumes that online lenders and a federal agency lender remain covered, while the other categories are largely removed by the new boundaries. That is a modeled composition, not an observed census of the market.
Rural coverage provides an even sharper test. Using branch locations as a proxy, the CFPB estimates that about 25% of rural depository branches would belong to institutions covered by the 2026 rule, compared with approximately 65% to 68% under the baseline. For non-rural branches, the estimated comparison is about 63% under the final rule versus 84% to 85% under the baseline. These figures do not show the share of rural borrowers who will actually lose access to credit or the share of branches that make small-business loans. They show how the rule’s institutional threshold may reduce the number of rural locations connected to reporting institutions.
The Bureau itself lists important limitations. There is no comprehensive market-wide dataset of small-business originations. CRA data cover some banks but not all, and the Bureau simulated plausible originations for institutions below the CRA reporting threshold. NCUA data provide no borrower location. A branch in a county does not prove that the institution serves every small business there, and the absence of a branch does not prove that it does not. The estimates are useful for comparing the agency’s baseline and final-rule models, but they cannot establish the final rule’s real-world effect on credit availability, prices, or discrimination.
The coverage design therefore embodies a specific trade. The CFPB is willing to lose a large number of reporting institutions and some dollar, rural, agricultural, alternative-finance, and growth-stage visibility in exchange for a first population it believes can report more consistently. Critics see the same design in the opposite order: the excluded lenders and products may be the places where borrowers have fewer alternatives and where fair-lending data are most needed. The sources do not resolve that dispute with post-implementation evidence.
What the rule does establish is a set of boundaries that can be evaluated. A complete assessment will need to ask not only how many applications enter the system, but also whose applications they are, which lenders and products disappear, how much dollar volume is represented, and whether local markets remain legible. The next question moves inside the applications that pass those gates: which facts will be recorded, which will be dropped, and what kind of oversight becomes possible when the data payload itself is smaller?