Leverage on Assets That Have No Price

2. Three Levels of Trust

Accounting standards (ASC 820, IFRS 13) sort every financial asset by how observable its price is.

Level 1. Trades on an active exchange. Price is the last trade. No models. No discretion. Stocks, treasuries, ETFs.

Level 2. Does not trade actively but similar assets do. Models calibrated to observable data: yield curves, comparable spreads, broker quotes. Corporate bonds, interest rate swaps. Constrained by real data.

Level 3. No market price. No comparable trades. Internal models with inputs the entity itself selects: discount rates, default assumptions, recovery estimates. Private credit, PE fund interests, CLO equity, real estate debt, infrastructure. The number is what the manager says it is.

Every illiquid asset a bank lends against is Level 3. Every tokenized RWA that is not a treasury bill or a stablecoin is Level 3.

3. Why Level 3 Assets Have No Price

A $1B syndicated loan gets structured with standardized documentation and distributed to 20 holders. Those pieces trade on a secondary market doing $800B to $1T in annual volume. Enough activity to produce a daily index.

A $75M direct lending loan sits with one fund. Bespoke agreement. Transfer requires borrower consent or is prohibited entirely. No trading desk. No dealer bids. If the lender wants out, it shops the position for months at a 5 to 15% discount.

The borrower chose this. One relationship, confidentiality, flexible terms. It pays 100 to 200 bps more for the privilege. The lender chose it too. Fatter spread for illiquidity. Same logic across Level 3: PE lockups, property specific real estate debt, bespoke infrastructure cash flows, complex CLO waterfalls. All built to not trade.

Then someone wants leverage on top. The bank says fine, but I need a price. There is no price. So the bank borrows one. Valuation firms pull daily syndicated loan spreads from the Morningstar LSTA index and PitchBook LCD, and apply them sideways to the private position that has never traded. A BB rated private loan probably moves in the same direction as BB rated syndicated loans. Probably. That is the entire basis for the mark.

4. How Banks Lever Up Against It

Total return swaps. Bank holds the asset, passes full economics to the fund. Fund pays SOFR plus a spread, posts 10 to 50% initial margin. The rest is leverage. Uses ISDA documentation. Dominant for single assets or small portfolios.

Repos. Fund sells asset to bank, buys it back at a higher price. Safe harbor under US bankruptcy law. Common for CLO tranches and bonds.

NAV facilities. Credit line against portfolio net asset value. $100 to $150B outstanding, projected to $600B by 2030.

Subscription lines. Secured by LP capital commitments, not assets.

Warehouse facilities and CLOs. For large diversified portfolios. The warehouse finances the ramp. The CLO securitizes the pool.

All five hit the same wall. The collateral was built to not have a price, and every one of these structures needs one.

5. Who Picks the Number

In a total return swap the bank is typically the Calculation Agent under ISDA. It financed the asset, holds it, and determines its value for margin purposes. If it marks the asset down, the fund gets a margin call. If the fund cannot post, the bank terminates and uses its own mark for the close out.

The conflict is structural. Mark conservatively during stress, trigger margin calls that protect the bank. Mark generously during calm, keep the trade alive and the fees flowing.

Funds negotiate valuation waterfalls: observable price first, then a dealer poll, then a model, then an independent valuation agent. But the less liquid the asset, the more the bank dictates terms.

The underlying method is the same everywhere. Valuation firms borrow price signals from public markets and apply them to private positions quarterly. The mark arrives 60 to 90 days late and captures broad market movements but nothing about the specific asset's behavior between reporting dates.

6. When It Breaks

For Level 1 collateral, the margin call is arithmetic. For Level 3, it is a dispute. The bank says 85 cents. The fund says 95. The difference is $50M on a $500M position. The fund gets 2 to 5 days to post. The asset takes months to sell.

Margin calls operate in days. Asset liquidation operates in months. That gap is where funds die.

7. What the Spread Tells You

The spread over SOFR is a direct measure of how much the bank distrusts its own valuation.

The bank cannot see the real price so it charges more and demands more collateral. The fund pays because leverage amplifies returns. A 10% gross return becomes north of 25% at 2.5x leverage.

7b. What the Loss Data Actually Shows

The largest specialist lender in this space publishes a dataset of 680 private equity funds from Preqin. The sample covers North American and European buyout funds with more than $100M in committed capital, vintage years from 1990 onward, all fully liquidated with final performance data available. The results:

At 25% loan to value, a lender loses money only if the fund returns less than 0.25x on its invested capital. In the entire 680 fund sample, the number of funds that lost 75% or more of their value is close to zero. At 10% loan to value, the fund would need to lose over 90% before the lender is impaired. That has functionally never happened in the sample for a diversified buyout fund above $100M.

The same dataset includes a stress test covering the Global Financial Crisis. A loan originated at 20% loan to value in December 2007, the peak before the crash, would have seen its effective LTV rise to 27% at the March 2009 trough as the PE index fell roughly 27%. By December 2010 the ratio had reverted to 20%. The loan was never close to impairment. The worst drawdown in modern private equity history moved the LTV by 7 percentage points on a 20% starting position.

Only 3% of funds in the sample returned less than half their invested capital. The median returned 1.7x. Yet the market standard advance rate is 5 to 25%. At 25% LTV, the fund would need to lose 75% of its value before the lender takes any loss at all. The gap between what the historical performance data supports and what lenders are willing to offer is the opacity premium in its most measurable form.

The lender is not being conservative because the assets are risky. The lender is being conservative because the assets are invisible. The NAV is a quarterly model output produced by the same manager who earns fees on it. Between marks, a portfolio company can deteriorate and the lender will not know for months. The low LTV is not a response to credit risk. It is a response to the information gap between quarterly reports.

8. The Price of Not Knowing

Go back to the spread table. A bank financing listed equities charges SOFR plus 75 bps and asks for 6% margin. A bank financing private credit charges SOFR plus 275 bps and asks for 30% margin. Same bank. Same balance sheet. Same legal infrastructure. The only difference is whether the bank can see what it is lending against.

The spread is not purely a credit risk premium. It is an opacity premium. The bank cannot see redemption behavior, leverage health, liquidity reserves, or who holds the position. So it charges more, demands more margin, and builds in wider termination rights. The entire cost structure of illiquid leverage is priced off the assumption that these inputs are unobservable.

Level 3 is not a property of the asset. It is a property of the infrastructure the asset sits on. A private credit loan is Level 3 because it sits inside a fund structure that publishes quarterly and reports through an administrator 60 to 90 days late. The credit risk did not make it Level 3. The infrastructure did.

Move the same exposure into a system where the fund's behavior is tracked automatically. The loan does not change. But redemption requests become visible when they are submitted. Supply changes are logged. Collateral ratios update continuously. Capital inflows are visible as they happen. Holder concentration is public.

A bank lending against a tokenized fund where it can see redemption queue depth, collateral health, and holder concentration in real time has more visibility than it has into most Level 2 assets. The margin buffer it needs is smaller. The spread it charges should compress. The valuation disputes that blow up funds become less frequent because both sides are reading the same contract state.

The entire cost structure described in sections 4 through 7 is calibrated to a world where the lender cannot see what is happening between quarterly reports. When that assumption breaks, the economics of illiquid leverage reprices.

9. Ravariant

Proxy baskets from the same public market instruments the valuation industry already relies on. Daily, transaction based data from markets that actually clear. That captures systematic risk: what the broad market is doing to assets with this profile.

On chain contract state that captures idiosyncratic risk: what this specific asset is doing right now.

Traditional valuation gets the first input daily and the second quarterly. Ravariant gets both continuously.

Two prices, on chain. Accounting value: what the asset is worth if the borrower pays and nothing forces your hand. Liquidation adjusted value: what you get if you need out now. Published transparently. Any protocol picks the number that matches its exposure. When this infrastructure is standardized and reusable, the cost of underwriting each loan drops and the savings pass through to borrowers.