Research Report
Bank and business surveys identify different constraints on smaller American firms. Working-capital accounts distinguish payment delays, collateral, repayment capacity, and the conditions under which guarantees improve financing.
The financing problem of a smaller business is rarely described adequately by the quantity of bank loans outstanding. A firm may be expanding its order book while losing access to credit because customers are paying more slowly. Another may reduce borrowing because its investment opportunities have deteriorated. A third may obtain the full amount requested only after accepting an expensive repayment schedule. These outcomes have different implications for employment, productive investment, financial stability, and the design of public support.
Dated evidence from the Federal Reserve's employer-firm survey and its bank lending survey indicates why these distinctions matter. The former observes financing experiences among businesses; the latter records changes reported by lenders. Neither independently identifies the national amount of economically viable investment prevented by credit constraints. This report combines those institutional observations with explicit lending and working-capital accounts. The analysis establishes where collateral monitoring, relationship information, and guarantees can improve financing, and where additional debt merely postpones the recognition of an operating loss.
The principal finding is conditional. An effective response depends on whether the binding constraint is payment timing, pledgeable assets, uncertain repayment capacity, or the lender's own resources. Lower interest rates address only part of that problem. Public guarantees can alter loss allocation, but their additionality requires evidence that finance, investment, or continuity improved relative to a credible alternative. A larger loan book alone cannot establish that result.
A useful definition of a financing gap begins with the activity that cannot be financed on available terms. The activity might be a seasonal purchase, delivery against a confirmed contract, replacement of machinery, or expansion into a new market. It must then be asked whether the expected cash generated by that activity supports repayment after allowing for risk. A difference between the amount requested and the amount offered is evidence of an unmet request. It becomes evidence of a socially consequential constraint only after the quality of the proposed use and the available alternatives are examined.
This distinction does not imply that lenders always assess projects correctly. Small firms may possess information that cannot easily be verified by outsiders. A customer relationship may be commercially valuable without being transferable collateral. New products may produce uncertain returns that are substantial in expectation but unsuitable for a fixed repayment obligation. These circumstances create potential financing failures. They also explain why equity, retained earnings, staged investment, and credit are imperfect substitutes rather than interchangeable quantities of money.
The relevant counterfactual is therefore more specific than unlimited lending at a lower price. Would another lender finance the same project? Would the business invest later using retained earnings? Would the owner contribute equity or reduce distributions? Would a supplier extend payment terms? Some alternatives are costly, and delay can destroy an opportunity. Nevertheless, ignoring them overstates the incremental activity associated with a supported loan. Conversely, treating the existence of an alternative as proof that finance is unconstrained understates the cost of obtaining it.
There is a further distinction between the firm and its owner. An owner may have liquid assets but reasonably refuse to pledge a family home against a commercial expansion. A loan can be available in a contractual sense while carrying a concentration of personal risk that deters investment. That is an important economic margin. It does not automatically justify transferring the risk to taxpayers, since the project's expected social return and the distribution of benefits remain separate questions.
The Federal Reserve Banks' 2026 employer-firm report, based on a survey conducted in September–November 2025, reports that 60% of respondents sought financing during the preceding twelve months. Among applicants, 42% received everything requested, 36% received part, and 22% received nothing. Operating expenses were a more frequently reported motive than expansion. These findings establish substantial variation in financing experiences; they do not make every unfilled request a viable investment forgone.
The survey covers employer businesses with fewer than 500 employees and is a weighted convenience sample. Its methodological documentation explains that participation comes through partner networks and direct outreach, with weights matching selected firm and owner characteristics. Question branching changes the denominator across items. Weighting improves the alignment of observed characteristics; it does not turn unobserved selection into random sampling. National extrapolations of a dollar financing gap would require additional information about requested amounts, business viability, and participation.
The distinction between applicant outcomes and population outcomes is especially consequential. Firms that already have adequate resources often do not apply. Others may be discouraged by expected rejection, documentation costs, or the personal assets required. Pooling those groups creates a misleading measure of unmet demand. An informative dataset would preserve the reason for nonapplication, the intended use, the terms anticipated, and whether the business subsequently financed the activity through another channel.
The Federal Reserve's April 2026 bank survey, Table 1, uses a different size boundary: annual sales below $50 million. Five of 61 respondents tightened small-firm lending standards and one eased them, a net balance of about 6.6 percentage points. Most reported no change. Meanwhile, nine narrowed small-firm loan spreads over funding costs and four widened them. Standards and price terms can therefore move differently within the same survey.
These are counts of responding banks, not weights based on loans or borrower employment. The questions concern changes over three months, rather than the absolute difficulty of obtaining credit. An unchanged standard can remain restrictive. The questions about loan terms concern applications the lender is willing to approve, excluding businesses screened out earlier. Taken together, the two surveys support a differentiated account of credit conditions. They do not support the assertion that every smaller American business faced either a uniform credit squeeze or uniform easing.
A business can earn an accounting profit and still run out of cash. Production requires purchases and payroll before the customer settles an invoice. The resulting financing need depends on the time between payments to suppliers and receipts from customers, the volume of activity, and the contribution margin. A larger order book can increase that need before it increases retained earnings. Credit demand arising from this sequence should be distinguished from borrowing used to cover repeated losses on ordinary sales.
Consider an illustrative distributor with annual sales of $3.6 million and purchases costing $2.4 million. Assume sales and purchases occur evenly over a 360-day commercial year. Customers pay after 45 days, inventory is held for 60 days, and suppliers provide 30 days of credit. These are constructed assumptions, not observed industry averages. Receivables equal $450,000, inventories valued at purchase cost equal $400,000, and payables equal $200,000. Net operating working capital is therefore $650,000 before cash buffers and other accruals.
If sales and purchases expand proportionately by one fifth while payment intervals remain unchanged, the corresponding requirement rises to $780,000. An additional $130,000 is absorbed by working capital. A business earning more on each completed sale can consequently require substantial external funds during expansion. The relevant underwriting question is whether invoices will convert to cash as assumed and whether the increased level of working capital is temporary, recurrent, or permanently larger.
The risk changes if customers extend payment from 45 to 60 days at the original scale. Receivables increase by $150,000 without any increase in sales. This looks like additional financing demand, but its interpretation depends on why customers are paying later. A contractual extension granted to strong customers may be manageable at an appropriate price. Widespread delinquency may signal worsening demand or customer solvency. A bank that treats those circumstances identically can either refuse sound activity or finance an accumulating loss.
| Constructed operating account | Initial business | Proportionate expansion | Slower collection, initial scale |
|---|---|---|---|
| Annual sales | $3,600,000 | $4,320,000 | $3,600,000 |
| Annual purchase cost | $2,400,000 | $2,880,000 | $2,400,000 |
| Receivables | $450,000 | $540,000 | $600,000 |
| Inventory at cost | $400,000 | $480,000 | $400,000 |
| Trade payables | $200,000 | $240,000 | $200,000 |
| Net operating working capital | $650,000 | $780,000 | $800,000 |
Receivables equal annual sales multiplied by collection days divided by 360. Inventories and payables use annual purchase cost multiplied by their respective holding or payment days, also divided by 360. The final row is receivables plus inventory less payables. Taxes, wages accrued, depreciation, cash balances, and capital expenditure are omitted deliberately. This exhibit isolates the trade cycle; it is not a complete forecast of cash available for debt service.
A lender evaluating the expanding distributor should also examine concentration. A receivable from one customer can dominate repayment even when the distributor has many invoices. Exposure to a customer's customer may matter if sales depend on a single construction project or retailer. The financing of apparently diversified small firms can therefore remain correlated at bank level. Firm counts alone understate that common exposure.
A revolving facility may specify a maximum commitment while limiting actual drawings to a borrowing base. The Office of the Comptroller of the Currency's asset-based lending handbook, particularly its overview and borrowing-base discussion, describes lending against eligible assets and the importance of monitoring their conversion to cash. The contractual ceiling and usable availability are distinct. Neither should be inferred from the balance outstanding on a reporting date.
An illustrative facility has a $500,000 commitment. Eligible receivables of $400,000 support an assumed 80% advance, and eligible inventory of $200,000 supports an assumed 40% advance. A reserve of $50,000 leaves a borrowing base of $350,000. If the borrower has already drawn $300,000, additional availability is $50,000, rather than the $200,000 suggested by the commitment less outstanding debt. The assumed advance rates are analytical inputs, not regulatory requirements or claims about typical contracts.
Now suppose delayed invoices reduce eligible receivables to $300,000, while inventory eligibility, advance rates, and the reserve remain unchanged. The borrowing base falls to $270,000. The existing drawing exceeds that amount by $30,000. A problem in customer collections has simultaneously increased cash need and reduced collateral-supported finance. The business may be required to repay when it is least able to do so. An interest-rate reduction would improve interest expense but would not automatically cure this availability deficit.
This mechanism helps explain why nominally ample commitments can coexist with a financing shortage. It also shows why unrestricted lending is an incomplete remedy. If the excluded invoices are ultimately uncollectible, advancing against them transfers a credit loss rather than bridges a timing gap. If they are temporarily late but sound, overly rigid eligibility rules can force unnecessary contraction. Effective monitoring must distinguish the quality and timing of the assets, which requires operational information rather than a single credit score.
The policy margin is consequently narrower than a broad instruction to increase credit. Better invoice verification, reliable financial reporting, and proportionate monitoring can make an otherwise uncertain repayment stream more financeable. Their value should be compared with their cost, particularly for small facilities. A sophisticated reporting arrangement that is economical for a large distributor may consume the expected benefit of a modest loan to a local service business.
Underwriting contains costs that do not increase proportionately with the amount lent. Understanding ownership, verifying tax records, evaluating customers, and obtaining enforceable security require time even for a small loan. A lender can therefore regard a small facility as uneconomic despite an acceptable probability of repayment. The spread required to cover those fixed costs may be high relative to a larger facility, without implying that the borrower is intrinsically less productive.
Repeated interaction can reduce the information problem. A bank observing transaction accounts may learn how receipts vary through the year, whether payroll is maintained, and how quickly the owner responds to an emerging shortfall. That information can make a business more understandable than a balance sheet alone suggests. It is not infallible: the same familiarity can encourage excessive accommodation, delay recognition of losses, or concentrate local exposures. Relationship lending has both informational benefits and governance requirements.
Small and large lenders face different operational choices, but institution size does not establish the causal quality of a loan decision. The employer-firm survey's higher full-approval rate among small-bank applicants is conditional on who applies there. Established firms with an existing relationship may select small banks, while urgent or difficult cases may apply elsewhere. Comparing approval percentages without matching applicants can attribute differences in borrower composition to differences in lending technology.
A defensible comparison would follow the same applicant across potential lenders or exploit a change in lender access that is unrelated to the business's own prospects. Applications should be matched by requested amount, purpose, financial strength, geography, and prior relationship. Even then, the terms offered must be compared: approval with a small limit or personal guarantee can be less attractive than a larger unsecured facility. Approval is a useful outcome, but it is only one dimension of access.
Digital underwriting may reduce processing costs and shorten response times. Its economic contribution depends on whether additional information improves repayment assessment, rather than merely automating a conventional rejection rule. Transaction data can be particularly useful for businesses with limited fixed assets. It can also be vulnerable to seasonal misclassification, changes in payment providers, and a short history that excludes severe downturns. Faster decisions should therefore be evaluated alongside the stability and explainability of those decisions.
A stated interest rate is insufficient to compare products whose fees, amortization patterns, and repayment contingencies differ. An installment loan amortizes principal; a revolving line may be drawn intermittently; a product repaid through frequent deductions from receipts can reduce the cash available for operations precisely when sales weaken. The economic cost depends on the amount of usable funding and its timing, including origination charges and required balances.
A simple hypothetical shows the importance of denominator choice. Suppose a firm receives $100,000 and repays $110,000 after one year, with no intermediate payments or fees. The annual cost is 10%. If $10,000 is withheld at origination while the same $110,000 is due, usable proceeds are $90,000 and the one-year cost becomes approximately 22.2%. Neither calculation describes a particular product. The comparison isolates the effect of fees on effective borrowing cost before introducing an actual dated schedule.
Frequent repayments require a cash-flow calculation rather than dividing a headline fee by an original principal. The outstanding balance declines through the year, so an identical dollar charge can correspond to a higher annualized financing cost. For commercial products, contract definitions and applicable disclosure rules also vary. A research comparison should publish dated cash flows and the method used, avoiding a confident rate estimate based on incomplete repayment information.
Affordability has a second dimension: the resilience of operating cash flow. A loan that is manageable under average annual receipts may create a seasonal deficit if payments are fixed while receipts fluctuate. A longer maturity reduces current payments but can leave debt outstanding after the financed asset is obsolete. Rescheduling may solve a genuine mismatch or conceal deterioration. The appropriate assessment follows the asset, the revenue it generates, and the timing of repayment together.
The employer-firm evidence on higher-than-expected online borrowing costs warrants attention to that process, rather than a conclusion that all nonbank credit is inferior. Different lenders may serve different risks and provide different speed or flexibility. The relevant comparison asks what the same borrower receives, what can change under the contract, and whether the initial decision was based on an accurate understanding of repayment obligations.
The Small Business Administration's 7(a) program description specifies lending through participating lenders, rather than ordinary direct borrowing from the agency. Eligibility includes creditworthiness and reasonable repayment ability, alongside inability to obtain the desired credit on reasonable terms through the specified nongovernment alternatives. The agency's description also identifies monitored working-capital facilities. These institutional requirements distinguish a guarantee intended to overcome an access problem from an unconditional subsidy for any requested expansion.
A guarantee changes who bears eligible credit losses; it does not create sales or collateral. Consider an explicitly hypothetical $200,000 loan with a 5% default probability and a 40% loss rate conditional on default. Expected credit loss before recovery costs is $4,000. If a guarantee covers three quarters of that eligible loss and all contractual conditions are satisfied, the lender's retained expected loss is $1,000 and the government's expected payment is $3,000. Administrative costs, fees, timing, and recoveries can change the fiscal result.
This arithmetic is an allocation of an assumed expected loss, not an estimate of the program's actual cost. It also does not establish that the guarantee causes a $200,000 increase in productive investment. The firm might have borrowed commercially, borrowed less, financed the purchase internally, or not undertaken it. To estimate additionality, an evaluation needs outcomes under those alternatives. The credit contract is the treatment; the amount disbursed is an intermediate variable.
Coverage can improve access where a lender understands the project but lacks sufficient collateral protection. It can be less effective where the principal problem is unreliable information, a weak market, or an inability to administer a small loan economically. If lenders retain little downside, screening incentives may weaken unless servicing and guarantee rules preserve meaningful responsibility. If conditions for payment are uncertain or expensive to document, the nominal coverage may be less valuable than it appears.
A strong evaluation would therefore track approval, terms, actual use, repayment, and business outcomes. Employment maintained during a temporary disruption differs from new employment generated by an expansion. Investment that replaces existing equipment differs from an increase in capacity. Outcomes should be measured over a horizon long enough to observe repayment and survival, without declaring every surviving recipient successful or every failed recipient evidence that support was inappropriate.
A bank must fund a loan, hold loss-absorbing resources, and manage exposures across its portfolio. A sound small-firm project can face tighter terms because the lender has lost deposits, faces higher wholesale funding costs, or is approaching an internal concentration limit. These are lender-side constraints. Their importance differs from a deterioration in the applicant's expected cash flow, even though the borrower may observe the same rejected application.
The distinction matters for monetary transmission. A lower market reference rate can reduce the benchmark component of a loan price while deposit competition, expected losses, or internal capital charges keep the total borrowing cost elevated. Conversely, a bank may narrow a quoted spread but reduce the maximum commitment or impose tighter collateral requirements. Looking only at a representative interest rate misses these quantity and eligibility margins. The April survey's separate questions are valuable precisely because they prevent those terms from being collapsed into one index.
A bank's constraint also need not be systemwide. A borrower may switch lenders if financial statements, collateral records, and a transaction history can be transferred at reasonable cost. Where those records are difficult to reproduce or collateral releases take time, an existing relationship can become a source of dependence. Switching costs make a local lending disruption more consequential, especially for a business that must pay suppliers before an alternative lender completes its review.
Public intervention should then be matched to the diagnosed mechanism. Information portability and timely lien administration address switching frictions. Liquidity assistance to an eligible financial institution addresses funding stress, subject to the relevant framework. A loan guarantee addresses a specified allocation of credit risk. Combining those instruments indiscriminately makes it difficult to establish which constraint changed and whether public resources improved the ultimate allocation of capital.
Falling loan growth can arise from supply, demand, repayment, write-offs, or changes in lender portfolios. A firm expecting fewer sales may cancel a planned expansion and draw less on its line even if credit terms are unchanged. Another may borrow more because receipts are weak. The direction of loan growth is therefore not a sufficient indicator of either confidence or constraint. Balance changes need to be linked to applications, commitments, drawings, and intended uses.
A practical empirical design would combine lender and borrower records. It would separate businesses exposed to banks experiencing an identifiable funding disturbance from comparable businesses using less affected lenders. Preexisting bank relationships should be fixed before the disturbance. Otherwise, firms may select a bank because of their own changing prospects. Industry demand, local conditions, firm age, and existing leverage require explicit treatment. The design should also test whether firms replace lost bank lending with other finance.
Substitution determines the economic effect. If another lender promptly finances the same sound project on similar terms, the original bank's contraction may change market share without reducing investment. If substitution is expensive or incomplete, investment and employment can respond even when total system lending remains stable. A national aggregate may conceal redistribution from constrained young firms toward established borrowers that are easier to evaluate.
There are competing explanations for apparently improved access as well. An approval rate can rise because weaker businesses stop applying or exit. Average borrower interest costs can fall because the remaining borrowers are safer. A guarantee program's participants can outperform nonparticipants because lenders select the strongest eligible firms. Each pattern is compatible with improvement, selection, or both. Observing the application pool before and after the change is essential to distinguish them.
The same caution applies to survival. Additional credit can keep a productive business operating through a temporary disruption, preserving customer relationships and trained teams. It can also maintain a business whose operating model no longer covers its costs. An assessment should examine post-support profitability, repayment, and market conditions, rather than equating continuity with efficient allocation. A credible evaluation permits both beneficial preservation and delayed restructuring as possible outcomes.
A borrowing need can be temporary while the proposed repayment obligation is excessive. To see the distinction, extend the distributor example with a separate, deliberately simplified income account. Assume annual operating cash generation before financing and taxes is $180,000. Existing scheduled principal payments are $80,000 and interest is $40,000. Cash remaining after those obligations is $60,000. The business can be profitable and current on its obligations while lacking the $130,000 required by its planned expansion in working capital.
Suppose the additional working capital is financed by a new $130,000 facility. Under an illustrative interest-only arrangement at 10%, annual interest adds $13,000 and leaves $47,000 after existing debt service. The principal still has to be repaid or refinanced when the facility matures. If the expansion permanently raises the working-capital requirement, the borrower cannot assume that routine collection of the first round of invoices will release all principal: the next round of sales requires replacement inventory and new receivables. A revolving asset is not necessarily a temporary funding requirement.
Under a different assumed arrangement, the same principal is repaid in five equal annual installments, with 10% interest on each year's opening balance. First-year principal and interest add $39,000, leaving $21,000. This comparison ignores taxes, loan fees, distributions, and any additional profit generated by expansion. It shows why repayment structure must be tested before interpreting approval as an economic improvement. The interest-only contract initially preserves more cash but retains refinancing risk; amortization reduces that risk by imposing a larger current claim.
The next step is to make the expansion's contribution explicit. If the additional $720,000 in annual sales generates $72,000 of incremental operating cash after its own recurring expenses, it can support more debt service than an expansion generating only $20,000. These are alternative assumptions, not estimated margins. The first case leaves room for repayment and contingencies; the second may produce a persistent shortfall once amortization and the enlarged working-capital stock are recognized. The size of the order alone does not distinguish them.
A stress case should then alter the variables that matter operationally. A payment delay affects both liquidity and borrowing-base eligibility. A reduction in the contribution margin affects debt-service capacity even when customers pay on time. Loss of a major customer may affect both. Combining every adverse assumption into one unqualified scenario is uninformative, but testing them separately reveals the channels through which a previously sound loan can become unsuitable. The exercise is more useful than applying a generic percentage reduction to all receipts and costs.
This account also clarifies the appropriate duration of support. A short bridge can finance an independently verifiable receipt expected after a delivery milestone. It is less suitable for a permanent increase in the capital tied up in ordinary operations. A guarantee with repeated extensions can become a continuing subsidy if the underlying business never generates enough cash to replace or repay the supported funding. Evaluation should distinguish a completed bridge, a renewed operating facility, and a loan that has merely avoided formal default.
For public policy, an exit condition should be specified when the support begins. Possible economic outcomes include access to an ordinary commercial facility, repayment from a completed contract, or accumulation of retained earnings sufficient to finance the recurring requirement. These are different paths and should be recorded separately. Requiring every firm to reduce debt on the same timetable can force contraction in a healthy growing business; permitting indefinite rollover without a repayment account obscures the actual fiscal exposure.
For a business leader, the first requirement is a forward cash account that distinguishes recurring operating losses from the financing of receivables, inventories, and investment. The account should identify the customer payments that service debt and show how delays change availability under the actual loan contract. A sensitivity to sales alone is inadequate if collateral eligibility deteriorates faster than sales. Concentration, timing, and repayment covenants deserve separate treatment.
For a lender, the useful question is which uncertainty can be reduced economically. A missing invoice record may be resolved through verification. An untested market may require equity rather than debt. A temporary concentration exposure may be addressed through participation or a smaller initial facility. A fundamental shortfall in operating returns cannot be eliminated by changing the label of the loan. This sequence makes underwriting more precise without requiring either universal expansion or universal retrenchment.
For policymakers, performance measures should follow the identified constraint. If a reform makes small facilities cheaper to process, monitor processing cost, time to decision, and borrower outcomes at a comparable risk. If a guarantee addresses collateral shortages, examine whether eligible firms obtain additional finance and undertake viable activity. If payment delays create the problem, measure collection intervals and the ability to finance confirmed receivables. Loan volumes can accompany these measures, but should not substitute for them.
Distribution also matters. A program can shift access toward firms whose owners have fewer pledgeable personal assets while leaving expected productive returns unchanged or improved. That may address a real inequality in the transmission of credit. Demonstrating it requires owner and business characteristics, comparable projects, and a clear account of who bears losses. It cannot be inferred solely from the number of loans made in a geographic area or to a designated category.
The evidence examined here supports a restrained but operational conclusion. Smaller American firms encounter several financing problems that are economically distinct and institutionally observable. Research improves policy when it specifies which one is binding, traces how the proposed instrument changes it, and measures the resulting activity against a credible alternative. The objective is finance that supports repayment-capable production and adaptation, with the costs and risks stated explicitly.
The official observations are dated: the employer-firm survey concerns 2025 experiences and was published in March 2026; the bank survey concerns changes reported in April 2026. They are not an assessment of every development through the information cutoff. The OCC handbook supplies institutional mechanics, while the SBA page identifies program design as inspected on that date. Neither source provides the numerical parameters used in the constructed accounts.
All distributor balances, borrowing-base calculations, effective-cost examples, and expected-loss amounts are hypothetical and reproducible from the stated assumptions. They establish mechanisms, not prevalence or estimated policy effects. The report does not estimate a national financing gap, an equilibrium loan price, or a causal employment multiplier. Those quantities require matched data and research designs beyond the published survey aggregates used here.