Section 04
The Cost of Capital
The Hook
Thesis
The sticker “SBA rate” is a fiction
This is price discrimination inside a government-branded product. The SBA 7(a) stamp implies uniformity: same guaranty, same agency underwriting box, same secondary market. The data shows a 220 bp spread on comparable deals (same industry, same year, same size, same rate structure) with no underwriting-risk explanation. The sticker “SBA rate” is a fiction. A lender-specific pricing policy exists, invisible at application, absent from every broker pitch.
Three charts, published annually: (1) rate dispersion scorecard by size and year, (2) prime-vs-spread pass-through across three conventions, (3) the two named lenders whose book-wide pricing fingerprints carry the dispersion story — Newtek Bank, NA (n=6,683 funded variable-rate 7(a) loans, FY23–25) and Readycap Lending, LLC (n=6,747). Both disclose broker-channel distribution. Neither shape looks like underwriting risk-pricing.
Framework
The Three Costs
Every borrower pays three costs. Only one shows up on the term sheet.
| Cost | What it is | Visible at signing? | Impact on $2M / 10 yr |
|---|---|---|---|
| 1. Spread | Margin over prime the lender charges. | Yes — rate is on the term sheet. | ~$200K swing between p10 and p90 lenders on comparable deals. |
| 2. Fees | Packaging, origination, broker-referral fees. | Partially. SBA guaranty fee disclosed; packaging / referral fees not in the public record. | $59K–$186K range at closing (Chart 8). |
| 3. Escape cost | Prepayment penalties + refi costs — the cost of exiting once the rate is locked. | No — buried in the promissory note. | 52% of $500K+ variable 7(a) loans pay off before half their term; pattern consistent with refi exit (Chart 5). |
Chart 1A
Rate Dispersion Scorecard
Variable-rate 7(a) Guaranty + FA$TRK, funded only, by size band and fiscal year. Cells are p10 / p50 / p90 of spread (initial interest rate minus prime at approval). (7(a), FY23–FY25, n=150,306, 7(a) Guaranty + FA$TRK, variable, funded (excl. CANCLD/NOT FUNDED))
| Size band | FY23 p10 / p50 / p90 (n) | FY24 p10 / p50 / p90 (n) | FY25 p10 / p50 / p90 (n) | FY25 p10–p90 (bps) |
|---|---|---|---|---|
| $0-$150K | P+1.50% / P+2.75% / P+5.80% (22,481) | P+2.00% / P+3.00% / P+6.50% (30,379) | P+2.00% / P+3.00% / P+6.49% (27,898) | 449 |
| $150K-$350K | P+1.00% / P+2.50% / P+3.00% (8,338) | P+1.00% / P+2.75% / P+3.49% (8,976) | P+1.50% / P+2.75% / P+4.50% (10,734) | 300 |
| $350K-$500K | P+0.50% / P+2.00% / P+2.75% (3,262) | P+1.00% / P+2.50% / P+3.00% (5,032) | P+1.00% / P+2.75% / P+3.00% (5,962) | 200 |
| $500K-$1M | P+0.75% / P+2.00% / P+2.75% (4,104) | P+0.75% / P+2.25% / P+3.00% (4,321) | P+0.75% / P+2.25% / P+3.00% (5,210) | 225 |
| $1M-$2M | P+0.50% / P+2.00% / P+2.75% (2,708) | P+0.50% / P+2.00% / P+3.00% (3,000) | P+0.50% / P+2.00% / P+2.75% (3,332) | 225 |
| $2M-$5M | P+0.25% / P+1.50% / P+2.75% (2,287) | P+0.25% / P+1.52% / P+2.75% (2,330) | P+0.14% / P+1.50% / P+2.75% (3,150) | 261 |
The smallest borrowers face the widest price range. A $0–$150K 7(a) borrower sits inside a 449-bpsspread between cheap and expensive lenders in FY25. A $2M–$5M borrower sits inside 261 bps. Every intuition says the biggest deals are where the negotiation happens, so dispersion should be widest at the top. The public data shows the opposite: the SBA market is least competitive exactly where borrowers have fewest resources to shop.
Chart 1B
Prime vs. Spread, Three Conventions
When prime moves, who captures the change? Three central-tendency conventions on funded variable 7(a) Guaranty + FA$TRK, FY21–FY25. All three agree: borrowers captured the full 91-bp pass-through from FY24 to FY25. (7(a), FY21–FY25, n=223,872, 7(a) Guaranty + FA$TRK, variable, funded (excl. CANCLD/NOT FUNDED))
| FY | n | Avg Prime | Vol-Weighted Spread | Simple-Avg Spread | Median Spread | p10 / p90 |
|---|---|---|---|---|---|---|
| FY2021 | 35,709 | 3.25% | P+1.90% | P+2.34% | P+2.25% | P+1.10% / P+3.00% |
| FY2022 | 34,659 | 4.10% | P+1.82% | P+2.44% | P+2.25% | P+1.00% / P+4.50% |
| FY2023 | 43,180 | 7.83% | P+1.80% | P+2.60% | P+2.50% | P+1.00% / P+4.75% |
| FY2024 | 54,038 | 8.48% | P+2.01% | P+2.96% | P+2.75% | P+1.25% / P+5.50% |
| FY2025 | 56,286 | 7.57% | P+1.99% | P+2.93% | P+2.75% | P+1.25% / P+5.00% |
From FY24 to FY25, prime fell 91 bps. Vol-weighted spread tightened 1.6 bps (1.994% → 1.985%) — statistically flat, in the borrower’s favor. Median spread held at P+2.75%. The median borrower’s rate fell a full 100 bps. The pass-through mechanism worked exactly as designed.
So why does rate shopping still save borrowers six figures? The pricing problem in 7(a) isn’t mean compression. It’s dispersion around the mean. Moving from the p10 lender to the p90 lender costs 375 bps (FY25). That’s more, over a 10-year loan, than the entire cumulative Fed relief of 175 bps since the cycle turned.
Chart 2
Industry Dispersion with Named Lender Extremes
Top-14 NAICS clusters by variable 7(a) volume, $1M–$3M, funded, FY23–25. Each row shows cluster p10 / p50 / p90 plus the named cheap-tier and expensive-tier lender (n ≥ 5 in bucket). (7(a), FY23–FY25, n=8,699, 7(a) Guaranty + FA$TRK, variable, funded, $1M-$3M, top-14 NAICS clusters by volume)
| Cluster | n | p10 / p50 / p90 | Cheap tier (n ≥ 5) | Expensive tier (n ≥ 5) |
|---|---|---|---|---|
| Restaurants & Food Service | 1,713 | P+0.75% / P+2.00% / P+2.75% | TD Bank, National Association (n=5, P−0.28%) | Newtek Bank, National Association (n=47, P+3.01%) |
| Miscellaneous Retail | 997 | P+0.75% / P+1.75% / P+2.75% | Mountain America FCU (n=6, P−0.70%) | Newtek Bank, National Association (n=17, P+3.04%) |
| Specialty Trade Contractors | 916 | P+0.75% / P+2.25% / P+3.00% | Mountain America FCU (n=7, P−0.87%) | Newtek Bank, National Association (n=39, P+2.99%) |
| Manufacturing | 868 | P+0.09% / P+2.00% / P+3.00% | Mountain America FCU (n=5, P−0.90%) | Newtek Bank, National Association (n=62, P+2.99%) |
| Hotels & Accommodation | 828 | P+1.00% / P+1.25% / P+2.25% | U.S. Bank, National Association (n=5, P−0.29%) | Newtek Bank, National Association (n=5, P+3.00%) |
| Food & Beverage Stores | 639 | P+1.00% / P+1.75% / P+2.75% | U.S. Bank, National Association (n=8, P−0.55%) | Newtek Bank, National Association (n=6, P+3.00%) |
| Wholesale Trade | 590 | P+0.25% / P+2.00% / P+3.00% | Comerica Bank (n=8, P−0.44%) | Newtek Bank, National Association (n=67, P+3.03%) |
| Administrative & Waste Services | 511 | P+1.00% / P+2.25% / P+3.00% | Enterprise Bank & Trust (n=5, P+0.45%) | Newtek Bank, National Association (n=32, P+3.01%) |
| Auto Repair & Maintenance | 497 | P+0.50% / P+1.75% / P+2.75% | America First FCU (n=5, P−0.87%) | Newtek Bank, National Association (n=9, P+3.00%) |
| Transportation & Warehousing | 478 | P+0.75% / P+2.25% / P+3.00% | U.S. Bank, National Association (n=10, P−0.07%) | Newtek Bank, National Association (n=5, P+3.00%) |
| Arts, Entertainment & Recreation | 470 | P+1.00% / P+2.25% / P+3.00% | Zions Bank (n=5, P+1.20%) | Newtek Bank, National Association (n=18, P+3.03%) |
| Physician & Dental Offices | 445 | P−0.50% / P+1.00% / P+3.00% | U.S. Bank, National Association (n=5, P−1.17%) | Newtek Bank, National Association (n=29, P+3.03%) |
| Social Assistance | 381 | P+0.35% / P+1.75% / P+2.75% | U.S. Bank, National Association (n=6, P−0.38%) | Newtek Bank, National Association (n=7, P+3.04%) |
| Real Estate | 366 | P+0.25% / P+1.35% / P+2.75% | U.S. Bank, National Association (n=5, P+0.23%) | Newtek Bank, National Association (n=11, P+3.02%) |
The same name keeps appearing at the expensive tier. Newtek Bank, NA is the named expensive-tier lender in 14 of 14 top industry clusters inside the $1M–$3M variable-rate 7(a) bucket (FY23–25, n ≥ 5 per bucket). The extremes aren’t random. They’re systematic, lender-specific pricing policies. See Chart 6 for the book-wide fingerprint.
Chart 3
The Signature $2M HVAC Example
Real data. Real lenders. Real math. NAICS 238220 (Plumbing, Heating, AC Contractors), $1M–$3M, variable 7(a) Guaranty + FA$TRK, funded, FY23–25. (7(a), FY23–FY25, n=229, 7(a) Guaranty + FA$TRK, variable, funded, NAICS 238220 (HVAC), $1M-$3M)
| Lender (deals in bucket) | Rate at P=6.75% | Monthly P&I | 10-yr total interest |
|---|---|---|---|
| Live Oak Banking Company (n=38) — P+1.43% | 8.18% | $24,456 | $934,739 |
| Newtek Bank, National Association (n=5, at Rule 5 floor) — P+3.00% | 9.75% | $26,154 | $1,138,486 |
| Delta | +157 bps | +$1,698/mo | +$203,747 |
Newtek’s n=5 at NAICS 238220 × $1M–$3M sits at the Rule 5 floor. The pattern is armored by the book-wide fingerprint in Chart 6: Newtek’s median spread is exactly P+3.00% in 90 of 93 (cluster × FY) buckets; 74.6% of its 6,683 funded variable-rate loans price at exactly P+3.00% to the basis point; 48.8% sit at or above cluster p90. The HVAC row is the screenshot version of a structural pricing policy, not a lone outlier.
Chart 4
Fixed vs. Variable: The Quiet Trap
$1.5M–$2.5M 7(a) Guaranty + FA$TRK, funded, FY23–25. (7(a), FY23–FY25, n=7,316, 7(a) Guaranty + FA$TRK, $1.5M-$2.5M, funded (excl. CANCLD/NOT FUNDED))
| Rate structure | n | Avg rate | Avg spread |
|---|---|---|---|
| Variable | 5,889 | 9.61% | P+1.70% |
| Fixed | 1,427 | 7.54% | P−0.37% |
Fixed-rate borrowers in FY23–25 got loans below prime. Variable borrowers paid prime + 170 bps. At origination, fixed looked 207 bps cheaper. 80.5%of $1.5M–$2.5M approvals went variable anyway. Variable is easier to sell into the secondary market; borrowers who don’t know to ask for fixed get the default, which is variable.
Chart 5
Prepayment: The Escape Rate
FY15–FY19 originations, $500K+ variable 7(a), observed through FY25 (mature enough to observe outcomes). (7(a), FY15–FY19, n=49,109, 7(a), >= $500K, variable rate, term_months not null)
| Spread at origination | n | % paid in full before half term | % charged off |
|---|---|---|---|
| P+<1.00% | 3,416 | 39.8% | 1.0% |
| P+1.00-1.99% | 16,680 | 51.9% | 1.4% |
| P+2.00-2.74% | 18,606 | 45.6% | 3.4% |
| P+2.75%+ | 10,407 | 41.7% | 5.4% |
Roughly half of $500K+ variable 7(a) loans pay off before half their term elapses — a pattern consistent with borrowers refinancing out of expensive pricing once the 3-year prepayment penalty burns off, though outcome data alone cannot attribute motivation. Chargeoff rates climb monotonically with origination spread: 0.97% at P+<1.00% vs 5.42% at P+2.75%+ — a 5.6x difference. Either exit takes the loan off the lender’s books; neither shows up in the stated amortization.
Chart 6 — Headline
The Broker Premium: An Open Question SBA Should Answer
The data gap is the story
The public record does not publish a broker-involvement flag. Every 7(a) record discloses the lender, the rate, the industry, and the borrower state — but not whether a broker originated the deal. That omission is load-bearing. Borrowers who believe they shopped their loan are rarely shown that their “lender” was really a broker’s preferred-buyer list; brokers whose referral fees run $10K–$30K per deal have no market-clearing incentive to disclose the counterfactual. The direct measurement exists in SBA Form 159 (Referral Fee disclosure) and SBA has chosen not to publish it. The proxy below is what we can publish.
Finding 1 — Newtek Bank, NA: the flat-ceiling pattern
Across 93(cluster × FY) buckets at n ≥ 5, Newtek’s median spread is exactly P+3.00% in 90 of 93 buckets. Book-wide: 6,683 funded variable-rate 7(a) loans; 74.6% are priced at exactly P+3.00% to the basis point. Measured against cluster × FY × size_band comparability buckets (n ≥ 5):
| Cluster position | Newtek | Live Oak (reference) |
|---|---|---|
| Below cluster p25 (cheap tier) | 0.25% | 38.05% |
| Cluster p25–p50 | 7.01% | 34.85% |
| Cluster p50–p75 | 25.80% | 14.53% |
| Cluster p75–p90 | 18.16% | 5.49% |
| At or above cluster p90 (most expensive decile) | 48.78% | 7.09% |
Read the row twice. Nearly half of Newtek’s funded book (48.8%) sits at or above the most expensive 10% of comparable SBA deals. Two-thirds (66.9%) sit at or above the most expensive 25%. Live Oak’s shape is the near-mirror: 72.9% of its 4,502 matched loans sit below cluster median. Same program, same guaranty, same secondary market, two lenders whose book shapes almost never overlap. The pattern is algorithmic pricing uncorrelated with borrower risk.(7(a), FY23–FY25, n=6,683, Newtek Bank NA, variable-rate 7(a) Guaranty + FA$TRK, funded)
Finding 2 — Readycap Lending, LLC: the step-function pattern
Cluster-level median spread jumped from P+2.75% in FY23 to P+6.50% across most clusters in FY24 — a 375-basis-point step function applied in the same fiscal year across every industry at once. FY25 median retraced to P+3.00% but mean remained elevated at P+4.47% (heavy right tail).
| FY | n | Avg spread | Median spread |
|---|---|---|---|
| FY2023 | 874 | P+2.79% | P+2.75% |
| FY2024 | 3,233 | P+4.92% | P+6.50% |
| FY2025 | 2,640 | P+4.47% | P+3.00% |
Book-wide cluster-position distribution (n_matched=6,740): 53.8% of Readycap’s funded variable-rate book sits in the most expensive 10% of comparable deals. That isn’t a statistical tail. It’s where the book lives. Two broker-channel-disclosed lenders. Two pricing fingerprints (Newtek’s flat ceiling, Readycap’s step reset). One outcome.(7(a), FY23–FY25, n=6,740, Readycap Lending LLC, variable-rate 7(a) Guaranty + FA$TRK, funded)
The upper-bound premium
Taking the spread gap between the broker-channel cohort (Newtek + Readycap + BayFirst) and the market median on comparable deals yields ~150 basis points. We report this as an upper boundon the broker premium, not a point estimate. It collapses three things the public record does not separate: actual referral-fee pass-through, credit-quality differences in the broker-channel cohort, and operating-margin differences between channel-dependent and direct-origination business models. On a $2M / 10-year 7(a) at 2025 prime, 150 bps of spread is $135,000 in lifetime payments — the same order of magnitude as a $10K–$30K at-closing referral fee capitalized and amortized over ten years. The arithmetic is plausible. The proof is in the Form 159 filings SBA has not released.
Chart 7
Restaurants: Different Industry, Same Math
The HVAC example (Chart 3) is one industry. To prove the gap is industry-independent we rerun the math on the largest acquisition vertical in the book: NAICS cluster “Restaurants & Food Service” at $1.5M–$2.5M, variable 7(a), FY23–25. Different named lenders at the extremes. Same structural gap. (7(a), FY23–FY25, n=721, 7(a) Guaranty + FA$TRK, variable, funded, $1.5M-$2.5M, NAICS cluster: Restaurants & Food Service)
| Lender (deals in bucket) | Avg spread | Role |
|---|---|---|
| Enterprise Bank & Trust (n=10) | P+0.75% | Cheap tier named lender |
| Live Oak Banking Company (n=25) | P+1.03% | Cheap tier (reference; see Chart 3) |
| Readycap Lending, LLC (n=30) | P+2.36% | Expensive tier (broker-channel disclosed) |
| Newtek Bank, National Association (n=17) | P+3.03% | Expensive tier (flat ceiling) |
Enterprise Bank & Trust (n=10) at P+0.75% vs. Readycap Lending (n=30) at P+2.36% on the same $2M restaurant profile. At 2026 prime, that’s $248,958 in extra interest over 10 years. Different industry than HVAC. Different named lenders at the extremes. Same structural gap.
Chart 8
Fees: What’s Disclosed, What’s Modeled
Only the SBA guaranty fee is statutorily defined and visible in the public record. Packaging, referral, and origination fees are modeled from SBA SOP 50-10 published ranges on a $2M 7(a) loan.
| Fee | Low case | High case | SOP cap / notes |
|---|---|---|---|
| SBA guaranty fee | $56,250 | $56,250 | Fixed by statute (2.75% on guaranteed portion above $1M in current SOP schedule) |
| Packaging fee | $2,500 | $30,000 | $30K SOP cap |
| Referral / broker fee | $0 | $60,000 | 3% of loan amount per SBA Form 159 |
| Origination fee | $0 | $40,000 | 0-2% lender-discretionary |
| Total out-of-pocket at closing | $58,750 | $186,250 |
On a $2M SBA 7(a), closing fees range from ~$59K to ~$186K. The difference is almost entirely who you used to get to the lender.
Methodology & data
See also
Where these numbers connect
- Section 1 — Scorecard: Chart D owns the prime-vs-spread teaser in its sidebar. The full longitudinal analysis lives here in Panel B.
- Section 2 — League Table: 514 bps mass-market dispersion across 151 top-100 lenders meeting the n ≥ 50 floor. Section 4’s 220 bps is the comparable-deal bucket (same industry, size, year, rate structure). Both are correct under their respective denominators. The Newtek row in Section 2 footnotes to this section.
- Section 3 — Money Flow: Live Oak Banking Company sits heavily in the acquisition-cohort volume; the cheap-tier discipline shown here shows up in their deal mix there.
- Section 5 — Borrower Playbook: The deal-killer patterns section uses the broker-premium fingerprint as one of four screening signals.