Advance Rate Methodology for Secured Lending Against Private Fund Interests

Scope

This methodology applies to non recourse loans collateralized by LP interests in closed end and open ended private funds: private credit, private equity, real assets, and infrastructure. Recovery is limited to the LP interest. No personal guarantees are taken.

It covers direct LP interests and tokenized representations (fund shares, feeder fund interests, and notes issued by securitization vehicles).

It does not apply to fund level NAV facilities, subscription lines, or loans against public securities. Those have established frameworks that differ from the approach here.

Key Pricing Drivers

Every loan turns on two questions: how much to lend against the NAV (loan to value), and what coupon to charge over the base rate. We pick the combination that compensates the lender fairly for the risks the loan carries, with strict limits on tail losses.

Five inputs are set at origination; four mechanisms run over the loan's life.

Outputs per loan: initial LTV, maintenance LTV, default LTV, coupon spread, cure periods, cash sweep triggers, stress sensitivity, liquidation path. The Outputs section details the schema.

Eligibility Criteria

Every gate is binary, applied to every asset, with no waiver.

Fund Eligibility

Monitoring Eligibility

Pool Level Gates (Ongoing)

The pool is every active loan inside the special purpose vehicle (SPV). Each loan is secured by a different LP interest, possibly in a different fund. The warehouse bank lends against the pool as a whole. Pool level rules prevent any single loan or fund from dominating the pool and protect the warehouse bank's senior position if individual loans go bad. We do not write loans that would break a pool rule.

Pass every gate and the asset moves on to advance rate determination. Fail any one gate and the asset is rejected. There is no override.

Data Quality Requirements (applies to all sections)

If a data source does not meet the standard, we throw it out. We do not give it a smaller weight; we exclude it.

Every input traces back to an independent, audited, or regulatory source.

Methodology Overview

The five inputs feed a single solver that picks the loan to value and coupon together, subject to the tail loss rule and the minimum return rule. Four other mechanisms run continuously while the loan is outstanding.

Three rules. First, each loan locks in its starting advance rate at origination and that number does not change for that loan; we may set a different rate for new loans against the same fund. Second, the dynamic margin threshold sits between the base rate and the 75% hard cap; it tightens when public stress signals flash, and a live loan to value above the threshold triggers an early margin call. Third, enforcement escalates step by step (warning, cash sweep, partial release, full release) on any confirmed covenant breach, not just loan to value. Full list in the Covenant Triggers section.

The base rate for new loans gets recalibrated when any of the following happens: rating change, the fund adds leverage, a redemption gate is exercised, fund AUM moves materially within our lookback window, a key person leaves, or the calibration interval expires.

Legal Structure and Loan Lifecycle

Entities and Agreements

Six entities participate in the system. Each pair is connected by a specific legal agreement that defines their obligations:

Repo Framing and No Recourse

Each loan is structured as a repurchase agreement: at origination the borrower sells the LP interest to the SPV, and at maturity buys it back for principal plus interest. The SPV holds title to the LP interest the entire time. Recovery is limited to the LP interest.

How a Loan Works: Step by Step

Repayment (No Default)

Full repayment triggers the repurchase leg. Contract verifies payment and releases tokens to borrower wallet automatically. No other party is required.

Operational Comparison

Base Advance Rate

The base advance rate is the percentage of NAV we lend at the start, before any live adjustments. We compute it in two steps: the formula below, then a strategy ceiling calibrated to NAV facility market practice.

The advance rate is applied to the LP's reported NAV, meaning the invested balance, not any unfunded commitment.

The Comparable Fund Set

The peer set comes from Preqin: 213,000+ funds, 64,900+ with performance data. We filter by strategy, AUM size, vintage year, geography, and leverage, and we keep only audited funds. Typical result: 30 to 300 funds with quarterly histories that we can compare against.

How We Measure Tail Loss (CVaR)

CVaR is the average NAV drop in the worst peer quarters. We use quarterly returns from Preqin and first apply Getmansky-Lo-Makarov unsmoothing (JFE 2004), an academic correction that reverses the artificial smoothness in private fund NAV reporting so the underlying volatility shows up. How deep into the tail we look depends on what the peer set actually lost. Three breaches in 40 fund years matches the 92.5th percentile; zero breaches means we go to the 95th percentile and add a 2 point cushion.

Splitting Tail Loss by Driver

The total tail loss can be split into contributions from each underlying driver (credit, rates, equities, fund specific) using a Gilchrist-Zakrajsek style regression. The point is to see what is actually moving the loss number, not just the total.

Strategy Ceiling

The formula handles market risk and structural risk but does not fully capture recovery uncertainty or the depth of the secondary market in a forced sale. The strategy ceiling caps the formula at NAV facility market practice. PIK loans (where unpaid interest gets added to the loan balance) get a lower ceiling because the loan grows even when NAV does not move.

The numerical ceilings are calibrated internally against KBRA NAV ratings, 17Capital practice, and the Fitch investment grade NAV cap. Whichever is lowest wins. Reviewed annually.

Maximum Acceptable Peer Loss Rate

Each asset class has a maximum loss rate it can show among its peers before we will not lend against it. The threshold comes from three things: how much we lend (the advance rate), how much loss the equity buffer below us absorbs first (the first loss tranche), and what we recover when we have to sell in a distressed market (Lazard secondary data).

Loss after recovery is the part of the loan we do not get back after a severely down position is sold in the secondary market. If we lend at advance rate AR, assume the NAV drops by d in the stress case, and Lazard data says we recover r of remaining value:

Worked example. If we lend 50% of NAV, the equity buffer below us is 20%, the safety factor is 50%, and we recover 70% in a distressed sale, then the loss after recovery is roughly 12% of the loan, and the bucket can show peer loss rates up to roughly 80% before we walk away. Reviewed annually.

The smaller the loan we make against a given NAV, the less of it we can lose when prices drop. So buckets with lower advance rates can tolerate higher peer loss rates and still pass. The binding constraint is advance rate, not the peer loss gate by itself. Recomputed quarterly.

Illiquidity Premium

Tail loss computed from past returns does not see things that only show up when you try to sell, like a long redemption queue or a freshly imposed gate. The illiquidity premium is an extra haircut for those. Each piece is calibrated either to secondary market data or to a known mechanical relationship.

The total adds up to a small percent of NAV in the cleanest case (liquid fund, no leverage, fresh mark) and rises into the high single digits or low double digits when pieces stack (fund leverage, prior gate, stale mark, large position). Exact bounds are internal. Recalibrated annually against Lazard and ILPA data. For context: performing LP interests in the Lazard secondary market trade at 92 to 94% of NAV, a 6 to 8% blended discount for credit and illiquidity together.

Manager Assessment

Larger, longer tenured managers tend to show smaller tail losses than smaller or newer managers holding similar assets. The manager modifier nudges the advance rate up or down to price that gap. For each dimension (size, experience, fund family) we compute the tail loss for the whole strategy peer set, then for just the funds matching that dimension, and take the difference.

The total is the sum across dimensions, with an internal cap so correlated strengths or weaknesses cannot pile on. Illustrative values from Preqin firm data for a representative direct lending bucket; production calibrations are internal.

Token Structure Assessment

When the collateral is a token instead of a direct LP interest, there is at least one extra entity between us and the fund: a securitization note, a feeder fund, a profit participating note, or some other wrapper. We adjust the advance rate and coupon for the extra structural risk. The adjustment is applied after the base rate, illiquidity premium, and manager modifier.

A direct LP holder owns equity in the fund and can take action in the fund's home jurisdiction. A token holder may own only a contractual claim against the wrapper, with no direct rights to the underlying fund. The wrapper is a separate entity with its own counterparty risk. Tokens with a documented redemption path to the underlying share keep the base rate; tokens with a blocked or uncertain path get three adjustments.

The underlying fund sets the best possible advance rate; the wrapper determines how far below that we end up. A feeder that gives the holder normal equity rights gets a mild adjustment. A note where the holder has no equity claim, can have the right wiped out if compartment assets run out, or cannot redeem on demand gets the full adjustment. Set at onboarding by reading the governing documents.

Dynamic Adjustment

Fund admins report NAV monthly at best, with a 45 to 90 day lag. The dynamic margin threshold fills that blind window using public macro indicators that have historically moved with fund credit quality. When those signals flash red, the threshold tightens before the next NAV report would have caught the stress.

Each indicator is converted into a standard score relative to its trailing history. The stress factor has three properties. It tightens fast and loosens slowly, so the threshold does not loosen on a single noisy reading. It uses a trailing average to tighten and a spot reading to loosen, which avoids whipsaw. It only ramps up when several indicators are in stress at once, not on a single noisy reading.

The threshold sits at or below the base rate, with a hard 75% cap above and an internal floor below. Indicator weights are set at onboarding under model governance.

Covenant Triggers and Enforcement

Two thresholds for loan to value. The dynamic margin threshold moves with stress signals and fires an early margin call when crossed. The 75% hard cap does not move and triggers liquidation; it is set against PIK NAV facility market practice and Lazard distressed recovery data. The base advance rate sets the loan size at origination; while the loan is live, we monitor it as (loan plus accrued interest) divided by the latest administrator NAV.

Two oracle systems feed the triggers. The NAV oracle ingests administrator NAV and cross checks against ARC, Preqin, and the factsheet before a 2 of 3 multisig push to the contract. The stress signal oracle feeds proxy basket signals into the dynamic threshold. A bad reading on either side does not become an enforcement decision until the other side confirms. This is the structural difference from a single price feed lending model.

Graduated enforcement: warning with cure, cash sweep, partial release, full release. Steps 1 and 2 are triggered by a dynamic margin call. Steps 3 and 4 require a verified breach. Automated via escrow contract.

Pool Diversification

Pool level concentration limits follow the Basel III Large Exposure Framework (BIS BCBS 283), which caps exposure to any single counterparty at 25% of a bank's Tier 1 capital.

Circuit Breakers

Origination halts and wind down begins on any of the following:

On hard stop: existing loans continue to maturity or are called on individual breach. All excess cash is trapped and applied to warehouse repayment. Monitoring and enforcement continue.

Limitations

The methodology covers systematic and structural risk; idiosyncratic risks specific to a given LP interest may remain.

Outputs and What We Do Not Publish

This is a pricing methodology, not a credit rating methodology. It describes how a lender (us) sets terms for its own loans. It does not assign letter grades to anyone else's debt. What this methodology produces, per loan:

What this methodology does not produce:

Specific calibrations, weights, peer set memberships, and the live proxy basket compositions are confidential. The framework above is sufficient to reproduce the methodology directionally; numeric values evolve under the change control on page 31.

Published under the publisher exclusion under U.S. securities law, the same legal posture a bank takes when it publishes its loan rate sheet or its underwriting standards.

Relationship to Rating Agencies

Rating agencies form opinions about whether the underlying assets will pay. We translate that into lending terms on loans secured by those assets. So their output feeds into ours. We start with the agency view as our best estimate of default probability and loss severity, then update it as our own pool generates real loss data. Early in the pool, the agency view dominates. As the pool matures, our own loss tape takes over.

If no rating exists, we start without one and rely on observable factors and our own pool experience. A new asset class does not stop us from lending; it just widens the safety buffer in the early advance rate.

Methodology Governance and Versioning

The framework is stable; calibrations, ceilings, and ranges evolve. Any prior pricing decision reproduces from the methodology version in effect at origination.

Change control and validation

Material changes (objective, risk constraint, asset universe, strategy ceiling) require Risk Committee approval, warehouse counterparty disclosure, and a change log entry. Validation follows Federal Reserve SR 11-7: Tier 1 components (pricing solver, NAV simulation, rating priors) receive annual validation and quarterly outcomes review; Tier 2 (default and loss models, proxy basket) annual and semiannual; Tier 3 (reporting) every two years. Unscheduled review on a Tier 1 outcomes exception, a material validation finding, a pattern of overrides, a regulator finding, or a loss breaching a published stress assumption.

Related Publications