Picture the conference room. Midtown Manhattan, thirty-eighth floor, floor-to-ceiling windows with a view that costs $180 per square foot per year. There are twelve people around the table. Four are from the structuring desk. Two are from legal. Three are from sales. One is from compliance and has the look of someone who has stopped sleeping. One is a consultant who has been describing himself as a "fintech bridge builder" since 2021. And there is a founder who used to run a merchant cash advance ISO and just raised a $40 million Series B on the promise that data science can make MCA underwriting not terrible.
The deck has forty-three slides. Slide three introduces the concept: Alternative Direct Revenue Receivables, or ADRRs.
The compliance person asks what an ADDRR is.
The founder explains that an ADDRR is a structured claim on a merchant's future revenue, originated through a proprietary underwriting framework that integrates real-time point-of-sale data, bank statement velocity analytics, and machine learning–derived repayment capacity scores.
The compliance person asks if that's an MCA.
There is a pause.
The consultant interjects that framing the asset class as MCAs would introduce unnecessary conceptual baggage for institutional allocators unfamiliar with the space, and that "Alternative Direct Revenue Receivable" more accurately captures the instrument's structural characteristics as a revenue-participation agreement rather than a debt obligation.
The compliance person writes something down. It is unclear what.
Slide seven introduces the fund vehicle: ACYG Fund I, LP — the Asset-Backed Securitization of Alternative-Credit Yield Generators. The target raise is $150 million. The target net return is 14 to 18 percent. The projected default rate is 6 to 8 percent. The consultant notes that this compares favorably to the high-yield bond index on a risk-adjusted basis.
The compliance person asks what the actual observed default rate is in the underlying portfolio.
The founder says that the term "default" doesn't quite apply to a revenue-participation instrument, and that the correct metric is "underperformance rate," which has been running at approximately 22 percent, though this includes a significant cohort of positions that are technically performing but behind schedule.
The compliance person puts the pen down.
Slide twelve introduces the waterfall structure, the senior and mezzanine tranches, and a brief mention of the equity tranche, which the deck refers to as the "Enhanced Return Participation Layer."
It used to be called the toxic waste. Someone in this building coined "enhanced return participation layer" at some point in the past, and now it appears on slide twelve of a deck describing MCA securitization, and no one in the room laughs.
This is where the humor ends. Because someone will build this. They may already be building it. And the history of what happens when Wall Street decides it has found a new asset class to repackage deserves a serious reading before the first ADDRR hits a Bloomberg terminal.
What Securitization Is, and Why It Works
Securitization is not inherently a scam. Before the industry learned to abuse it, it was a genuinely useful innovation in capital markets.
The basic mechanism: a financial institution — a bank, a lender, a specialty finance company — originates a pool of assets that generate cash flows. Mortgages, auto loans, credit card receivables, student loans. Those assets sit on the originator's balance sheet and tie up capital. Securitization allows the originator to sell the pool to a special purpose vehicle (SPV) — a legally distinct entity — which issues securities backed by those cash flows. Investors buy the securities. The originator gets its capital back and can originate more loans. The investors get the cash flows.
The SPV then structures those cash flows into tranches — slices of the pool with different seniority claims.
The senior tranche gets paid first. It absorbs almost no losses unless defaults are catastrophic. It's the safest. It earns the lowest return and typically receives an investment-grade credit rating. Conservative institutional investors — pension funds, insurance companies — buy senior tranches.
The mezzanine tranche gets paid after the senior. It absorbs losses once they exceed what the equity tranche can absorb. It's riskier, earns more, and gets a lower credit rating.
The equity tranche — the "first-loss position" — absorbs all losses before any other tranche. It's the riskiest. It can earn exceptional returns if defaults are low. It can be wiped out entirely if defaults are high. The originator often retains the equity tranche, which is supposed to "keep skin in the game" and align incentives.
This structure makes theoretical sense. Institutional investors with different risk tolerances can access the same underlying cash flows at their preferred risk level. The originator gains capital efficiency. The pool diversifies idiosyncratic risk — even if some loans default, the pool as a whole may perform fine.
The mechanism worked well for decades for vanilla mortgage pools, auto loans, and credit cards. Then two things happened: the underlying assets got worse, and the math used to analyze them got wrong in a catastrophically specific way.
The Formula That Broke the World
In 2000, David X. Li, a quantitative analyst at JPMorgan, published a paper introducing a method for modeling the correlation of defaults in a pool of assets. The technique used a mathematical tool called the Gaussian copula — a way to describe the relationship between two probability distributions, in this case, the likelihood that two assets would default around the same time.
The formula was elegant. It reduced the complex, interconnected question of "how will these assets fail together?" to a single number: a correlation coefficient, conventionally written as ρ (rho). If ρ is 0, defaults are completely independent — one asset failing tells you nothing about whether another will. If ρ is 1, defaults are perfectly correlated — if one fails, they all fail. Real portfolios, of course, fall somewhere between.
The finance industry adopted Li's model with near-universal enthusiasm. It allowed rating agencies, banks, and investors to price complex structured products — CDOs (Collateralized Debt Obligations), CDO-squareds (CDOs of CDOs), synthetic CDOs backed by credit default swaps rather than real loans — with mathematical precision and speed.
The problem was where the correlation inputs came from.
Because historical data on correlated mortgage defaults was limited — the United States had not experienced a nationwide simultaneous collapse in housing prices since the Great Depression — quants estimated ρ from credit default swap spreads in the market. This was circular: the market prices reflected models that assumed ρ was low and stable, so the model inputs confirmed what the model assumed.
The actual correlation between defaults in a mortgage pool — especially a subprime mortgage pool — was not low and stable. It was latent, conditionally high, and explosive when triggered.
Subprime mortgages were underwritten on the assumption that housing prices would continue rising. When a borrower defaulted, the collateral (the house) could be sold at a value that recovered most of the loss. This worked as long as houses kept appreciating. Default correlation didn't matter much because recoveries were high.
When housing prices stopped rising in 2006 and began declining in 2007, the entire assumption structure collapsed simultaneously across the entire country. Defaults didn't rise modestly and gradually — they surged in correlated waves. Loss-given-default rates, which the models had pegged at 20 to 30 percent based on historical sales of foreclosed houses in rising markets, moved toward 60 to 80 percent in falling markets with massive simultaneous inventory.
The senior tranches, modeled as safe because correlation was assumed low, turned out to be exposed to exactly the scenario the models said couldn't happen: mass simultaneous default. AAA-rated securities became worthless in months.
The correlation assumption was the flaw. And the correlation assumption was based on data from a period that was, by historical standards, anomalous.
Why MCA Securitization Is Worse
Subprime mortgages, for all their recklessness, had several properties that at least gave the math something real to work with. They had standard documentation. They had thirty-year fixed payment schedules. They had collateral — physical real estate. They had a century of performance data. Rating agencies had at least modeled something real, even if badly.
Merchant cash advances have none of these properties. And the ways MCAs differ from subprime mortgages make them more dangerous to securitize, not less.
No fixed payment schedule. MCAs repay as a percentage of daily or weekly credit card and bank revenue. Repayment accelerates when the business does well and decelerates when it struggles. This makes duration modeling nearly impossible — a nominal six-month advance can take eighteen months to repay if the merchant's revenue drops. Cash flow models built on "expected repayment velocity" are building on sand.
No standardized underwriting. Mortgage underwriting, for all its subprime failures, at least had consistent documentation requirements — tax returns, W-2s, appraisals. MCA underwriting varies enormously by funder. Some do three months of bank statements. Some do one. Some use factor rate tables that don't meaningfully differentiate between a restaurant with 80% margins and a grocery store with 2% margins. There is no FICO equivalent. There is no standardized default definition. An MCA that stops receiving payments may be "in default," or it may be documented as "seasonal slowdown" or "position under review" — depending entirely on what the funder decides to call it.
The stacking problem. A significant portion of MCA borrowers have simultaneous advances from multiple funders — a practice called stacking. Funder A provides an advance. The merchant uses a portion to pay daily remittances to Funder A. Months later, Funder B provides a second advance. The merchant now has two remittances coming out daily. Funder C follows. The merchant's daily cash outflow from remittances can exceed their operating margin. This cascades.
In a securitized pool, the interaction between stacked positions across funders creates correlation that no individual pool analysis would detect. Merchant A defaulting on Funder 1's advance may simultaneously trigger defaults on Funders 2 and 3's advances, which appear in different pools managed by different managers. The correlation isn't within the pool — it's across pools, invisible to any individual analysis.
Correlation to economic conditions is extreme and asymmetric. MCA borrowers are overwhelmingly small businesses in cash-intensive industries — restaurants, retail, hospitality, medical practices, salons. These businesses have highly volatile revenue that correlates strongly with macroeconomic conditions. In a recession, they don't underperform gradually. They close.
The correlation coefficient in an ADDRR pool isn't low. It's high, conditionally extreme, and concentrated in exactly the scenarios — economic downturn, credit tightening — in which investors in a structured product need it to be low.
Recovery is near zero. A defaulted subprime mortgage had a house behind it. In a terrible scenario, the lender got 40 cents on the dollar. An MCA claim against a closed restaurant is an unsecured claim against a defunct entity with no significant assets. Recovery rates in MCA default approach zero in the worst scenarios. Loss-given-default is not 40 to 60 percent. It is 90 to 100 percent.
Duration mismatch in a securitization vehicle. MCAs have nominal durations of three to eighteen months. A securitization vehicle, to attract institutional capital, typically targets a three-to-five-year fund life. This creates a reinvestment requirement: as MCAs repay (or default), the vehicle must originate new positions to maintain the portfolio. The securitization is not buying a fixed pool of cash flows — it's buying continuous exposure to a live origination machine. If origination quality deteriorates during the fund life — and in MCA, it typically does when originators are chasing volume to feed a fund — the pool's quality can drift substantially from what was presented in the offering documents.
The Tranche Math Applied to ADRRs
Let us perform the exercise. Suppose ACYG Fund I, LP originates a $100 million pool of ADRRs. The pool has the following structure:
- Senior tranche: $70 million (70%), targeting a BBB equivalent rating, sold to institutional investors at an 8% coupon
- Mezzanine tranche: $20 million (20%), targeting a B equivalent rating, sold at a 14% coupon
- Equity tranche: $10 million (10%), retained by the originator, targeting 22% net return
The offering document states that the expected default rate is 7%, with expected recovery of 15% (unusually optimistic given the near-zero recovery typical of closed-business MCA claims). Expected net loss: 7% × (1 - 0.15) = 5.95%.
With 10% equity absorbing first losses, the senior tranche is theoretically protected unless net losses exceed 10%. A 6% net loss scenario leaves the senior tranche fully intact. The rating case is made.
Now run the actual numbers from the MCA industry.
Default rates in MCA pools — variously defined, inconsistently reported — are documented across funders in the range of 15 to 25 percent under normal economic conditions. In the period from March through June 2020, when restaurant revenue dropped 60 to 90 percent in lockdown-affected areas, effective default rates in MCA portfolios concentrated in hospitality exceeded 40 percent.
Apply the stress scenario: 40% default rate, 5% recovery (a closed restaurant's bank account, less legal costs). Net loss: 40% × 95% = 38%.
The 10% equity tranche is wiped out in the first wave.
The 20% mezzanine tranche absorbs losses up to 30%. In a 38% loss scenario, the mezzanine is wiped out with 8 points of loss left over.
The $70 million senior tranche — the AAA-seeking, pension-fund-targeted, "diversified high-yield" position — takes an 8% loss. On a $70 million position, that's $5.6 million of losses in a tranche that was supposed to be safe.
And this analysis assumes the pool was honestly constructed, that default rates are accurately reported, that stacking exposure across the industry isn't concentrating correlated risk, and that the economic shock doesn't persist for more than a quarter. None of those assumptions held in 2020 and there is no structural reason to believe they will hold in the next downturn.
The Rating Agency Problem
For ACYG Fund I, LP to succeed, it needs a credit rating. Rating agencies — Moody's, S&P, Fitch — will be asked to analyze the pool and assign ratings to the tranches. This is what gives institutional investors the regulatory cover to buy senior positions.
Rating agencies charge issuers for ratings. This is the conflict of interest that structured finance critics have identified since the 2008 crisis. An issuer who doesn't like a rating shops for a different agency. Agencies competing for mandate revenue have an incentive to accommodate favorable requests.
But beyond the incentive problem is the competence problem. Rating analysts who cover corporate bonds, municipal debt, and traditional structured products do not have deep familiarity with MCA underwriting quality, MCA stacking patterns, MCA recovery rates in default, or the specific ways that revenue-based repayment accelerates correlation risk during economic contraction.
They will be handed a model. The model will be provided by the issuer, perhaps with independent verification from a third-party analytics firm whose fee is also paid by the issuer. The model will project default rates, recovery rates, and correlation coefficients. The correlation coefficient will be low because the historical data from which it was estimated did not include a meaningful economic stress period with concentrated MCA exposure.
This is the Gaussian copula problem, restaged in a new asset class, with rating analysts who are less familiar with the underlying assets than their mortgage-crisis predecessors were with subprime.
The Question No One in the Conference Room Asked
Return to the thirty-eighth floor. There is a question that does not appear in the forty-three slides and is not asked by anyone in the room.
The merchants whose advances are being packaged into ACYG Fund I, LP are paying — on an annualized basis — somewhere between 40 and 150 percent for capital. They're paying this because they cannot access conventional credit at a price that would allow them to survive. The pizza shop is paying a 1.45 factor rate on a six-month advance because its margins, its credit history, and its collateral make a bank unwilling to lend to it at a rate that would make the math work for anyone.
Securitization does not change this. Packaging those advances into a fund, tranching them, rating them, and selling senior positions to pension funds does not improve the pizza shop's situation. It does not reduce the cost of capital to the underlying borrower. It does not improve underwriting. It does not create accountability. It moves the risk from the originator's balance sheet to the fund's investors — and in doing so, removes the originator's incentive to care about whether the advances it sells into the pool are good ones.
This is the originator/distributor model that produced the subprime crisis. Originators who retain loans on their balance sheet care whether those loans perform. Originators who sell loans into securitization vehicles the moment after funding care whether the loan passes the initial underwriting screen to be eligible for sale. The two incentive structures are fundamentally different, and the second produces systematically worse underwriting.
The compliance person in the conference room is writing down something like this.
The other eleven people in the room are thinking about carried interest.
Wall Street will not fail to build this because it's a bad idea. Wall Street will build it because there is money to be made building it, distributing it, and managing it — and the people who will eventually hold the losses are not in the room.
They are in a pension fund in Ohio. They are in an insurance company's fixed income portfolio. And they are in every small business across America whose cost of capital will rise when the first vintage of ACYG Fund I, LP performs exactly as the math predicts, regulators notice, and the alternative credit market briefly seizes up while everyone in the conference room finds a new conference room.
The compliance person was right to stop sleeping.
For commercial borrowers navigating the MCA landscape without the benefit of a thirty-eighth-floor view: FundScout connects businesses to vetted lenders through a marketplace where the incentive structure is transparent. Lenders who accept a match pay for that relationship. They don't get paid to originate loans into a fund they'll never see again.
Sources
- David X. Li, "On Default Correlation: A Copula Function Approach", Journal of Fixed Income, Vol. 9, No. 4 (March 2000) — introduced the Gaussian copula to credit correlation modeling; Risk.net
- Felix Salmon, "Recipe for Disaster: The Formula That Killed Wall Street", Wired, February 23, 2009 — Li's formula and its misapplication in structured finance
- Financial Crisis Inquiry Commission (FCIC), The Financial Crisis Inquiry Report (2011) — govinfo.gov; analysis of CDO and subprime securitization failures; loss-given-default data
- Subprime default and loss severity data (2007–2010) — Federal Housing Finance Agency; Federal Reserve; loss severity in foreclosure reached 40–80% in falling markets vs. 20–30% historical
- MCA default rates (15–25% under normal conditions; 40%+ in hospitality during March–June 2020) — consistent with small business lending stress data from Federal Reserve Small Business Credit Survey 2020
- Anna Gelpern & Adam J. Levitin, "Rewriting Frankenstein Contracts", Southern California Law Review, Vol. 82 (2009) — originator/distributor incentive misalignment in structured finance
- Moody's, S&P, Fitch rating agency conflicts of interest — documented in FCIC Report, Chapter 11; Senate Permanent Subcommittee on Investigations, Wall Street and the Financial Crisis (2011)