Balance Sheet CLOs: Moving Risk Off the Books

Book: Structured Finance and Collateralized Debt Obligations | Author: Janet M. Tavakoli | Publisher: John Wiley & Sons (2008) | ISBN: 978-0-470-44344-6

Chapter 12 shifts focus from arbitrage CDOs to balance-sheet CDOs. The primary motivation is different. Banks are not trying to profit from the spread between asset yields and liability costs. They are trying to move credit risk off their balance sheets and free up regulatory capital.

The technology for this goes back to consumer loan securitization in the 1980s. Chapter 12 tracks how that technology evolved and branched into multiple structural forms, each with different trade-offs.

True Sale (Delinked Structure)

The first rated CLO backed by US bank loans came in 1990. The earliest CLOs were arbitrage CLOs focused on equity returns from purchased market collateral. But banks with loans already on their books had different goals: regulatory capital relief and off-balance-sheet funding.

In a true sale CLO, the bank originates loans and sells them to an SPV. The SPV issues notes to investors. The sale transfers both legal ownership and risk to the SPV. The bank is delinked from the investor, meaning the investor does not depend on the bank’s creditworthiness to get paid.

Cash flow from interest on the underlying loans services the deal liabilities: tranche coupons and fees. Excess spread goes to a reserve account.

These deals have a lockout period of two to four years. During lockout, the deal does not amortize. The manager reinvests principal from maturing loans into new loans per criteria in the deal prospectus. After the lockout, the deal amortizes on a sequential basis by tranche seniority.

Early termination can be triggered by insolvency of the issuer, breach of collateral maintenance boundaries, or reaching a threshold level of defaults.

Commercial and industrial loans can be bilateral (the bank is the sole lender) or multilateral (the loan is syndicated among several underwriters). By selling loans to an SPV, the bank reduces its balance sheet and can theoretically originate new business.

But Tavakoli notes something surprising: some financial institutions reject this structure because they do not want to reduce balance sheet size. German savings bank officers are compensated based on how large their balance sheet is. A major true sale CLO works against their personal incentives.

Interest rate hedging is needed for fixed-rate loans in a floating-rate CLO. CLO tranches pay LIBOR-indexed coupons. Fixed-rate loans need to be swapped. This introduces basis risk if the loan coupon spread is variable (many loans allow the spread to increase or decrease based on the borrower’s financial condition).

CLOs have ranged from $300 million to $5 billion in size, though larger deals tend to be synthetic.

For US FDIC institutions, the step of selling into an SPV is often not required to achieve bankruptcy-remote status. Figure 12.1 in the book shows how traditional ABS technology using an intermediate SPV and a master trust structure applies to a true sale CLO.

True sale documentation is intensive: purchase agreements for each note class, Reg S and Rule 144A global note documentation, paying agency agreement, trust deed, cash bond administration agreement, asset sale agreements, servicer agreement, and additional documentation if an interest subparticipation tranche is included.

The moral hazard concern in balance-sheet CLOs: once loans are sold, the bank may let portfolio quality slip since the loans are no longer on its books. The counterargument is that if the bank retains servicing rights and holds the equity tranche, reputational risk and economic incentive keep them honest. Tavakoli’s view: both schools of thought have some merit. Structural protections can mitigate but never eliminate this risk. You can protect against moral hazard, but not against incompetence or fraud.

Linked Nonsynthetic Structures

In 1996, National Westminster Bank brought the $5 billion R.O.S.E. (Repeat Offering Securitization Entity Funding No. 1) deal to market. NatWest retained legal ownership of the loans but transferred risk via subparticipation interests. The collateral was revolving corporate loan facilities.

This was a linked structure: investor returns depended partly on NatWest’s own creditworthiness. Without delinked status, the highest achievable tranche rating was NatWest’s own rating, which was Aa2/AA at the time. No AAA tranche was possible.

NatWest received regulatory capital relief under BIS I guidelines. The equity tranche was only 2% of the deal, so NatWest held $100 million of capital against it instead of $400 million before securitization. A 4:1 reduction.

But from a cost perspective it was inefficient. The Aa2/AA tranche made up 95.9% of the deal. If it had been delinked, that tranche would have shrunk to about 2%, and a new AAA tranche of about 94% would have been much cheaper to fund. The liability cost of the largest tranche increased by the full spread difference between AAA and AA2. Tavakoli quotes the deal’s nickname without apology: “A R.O.S.E. by any other name would still stink from the point of view of cost efficiency.”

Linked Black-Box CLN CDOs: The Glacier Deal

What if a bank wanted to transfer credit risk not just from its loan book but from its entire enterprise, including swap counterparty exposure and securities positions?

SBC (then Swiss Bank Corporation, rated AA+/AA1) answered that question on September 10, 1997, with Glacier. The deal was so popular it was upsized from $1.5 billion to $1.75 billion at launch.

Glacier Finance is a master trust SPE domiciled in the Cayman Islands. The portfolio consists of credit-linked notes issued by SBC itself. The CLNs embed credit risk from SBC’s investment and commercial banking businesses: loans, derivatives, securities, and project loans. The CLNs are callable by SBC on any interest payment date, and SBC can rotate credits in and out of the SPV.

The deal is also a black box. Swiss banking laws require high client confidentiality with up to 15-year prison terms for violations. To preserve confidentiality, SBC embedded client credit risks inside CLNs. Investors do not know the composition of the reference pool. SBC’s own relationship managers and credit officers do not know which client risks are in the SPV. Only a few key credit officers have access.

SBC mapped its internal ratings to Moody’s and S&P ratings to allow continuous collateral rotation. SBC determines credit events for the MTNs. If a credit event occurs, SBC calls the CLN and redeems it at recovery value, determined by the senior unsecured debt of the reference credit. If no reference security exists, payout is preset at 51%, the average recovery rate for senior unsecured obligations per Moody’s data at deal close.

At the time, SBC estimated a 5:1 regulatory capital reduction. The deal was a linked structure, but it sold well because there was not much else available for investors wanting access to large, diversified pools of bank portfolio risk.

The first Triangle deals had similar structures. Today, investors would not accept either a linked structure or a black box. But these were pioneering deals and the standards of disclosure were different.

Synthetic Structure with SPE: BISTRO

JPMorgan brought the first BISTRO (Broad Index Secured Trust Offering) to market in December 1997. It was the first CDO to fully exploit synthetic structure with an SPE.

For the first time, default risk on a pool of loans was transferred using credit default swaps rather than true sale. JPMorgan as the bank originator bought credit default protection on a diversified loan portfolio. The super senior risk was retained by JPMorgan (they liked it). The mezzanine risk was purchased by noteholders in the form of credit-linked notes issued by the BISTRO SPE. Note proceeds were used to purchase AAA collateral.

The SPE is delinked from JPMorgan’s credit risk. If JPMorgan had problems, investors in the notes were protected by the AAA collateral backing the notes, not by JPMorgan’s balance sheet. That was a significant advance over the linked structures.

The super senior swap counterparty is almost always a 20% BIS risk-weighted entity, either an OECD bank, a financial institution, or an agency. This counterparty may retain the risk or lay it off to a monoline insurer, reinsurer, or financial guarantor.

Rating agencies doing BISTRO-type deals examined the bank originator and assets using standard criteria: administration and organization of the bank, credit policies and controls, technology systems, decision-making processes for adding assets, and internal reporting. For the asset pool: obligor diversification including industry and regional concentration, liquidation profiles, delinquency and default data, and recovery value data.

The equity is almost always retained by the bank originator, which means they are not getting full regulatory capital relief on it. But they get partial relief: the equity is held on balance sheet at dollar-for-dollar capital, but the mezzanine and AAA tranches get lower capital treatment.

European variations sometimes incorporated interest subparticipation tranches for the equity portion, using the structure described in the previous chapter to capture favorable capital treatment.

Documentation for BISTRO-type deals is much lighter than for true sale CLOs: terms and conditions for each note class, credit default swap confirmations, and the trust agreement.

Partially Synthetic Linked CDOs

LB Kiel (Landesbank Kiel) brought a EUR 1 billion balance-sheet CDO to market in 2001 under the name FÖRDE 2000-1. The underlying collateral was first and second residential mortgage loans. LB Kiel was rated Aa1/A1+/AAA at the time.

LB Kiel transferred super senior risk via a super senior swap. For the other tranches, LB Kiel itself acted as issuer of credit-linked notes. The equity tranche was sold as a privately rated interest subparticipation.

To get AAA/AAA ratings on the AAA CLN, LB Kiel had to use AAA Pfandbriefe as collateral to back the principal repayment of the note. This was a partially synthetic linked structure using many of the most innovative elements available at the time.

Fully Synthetic CDOs

If you can find counterparties for all credit default swaps directly, you do not need an SPV at all.

In a fully synthetic balance-sheet CLO, the bank originator buys CDS protection tranche by tranche from different counterparties: the super senior from one, mezzanine from noteholders via an SPE, equity retained. The AAA tranche may require the AAA counterparty to post AAA collateral.

For synthetic arbitrage CDOs, the bank arranger’s CDS trading desk provides the reference assets rather than a bank’s balance sheet. The structure is the same mechanically, but the source of the reference portfolio is different.

Many investors want to take on tranched synthetic risk but need it in note form. Two ways to package it: a dedicated SPE for each deal, prefunded with limited recourse notes, or a rated multi-issuance SPE that can accommodate multiple deals with each tranche ring-fenced. The SPV buys AAA-rated collateral like Pfandbriefe to back a AAA credit-linked note linked to the AAA CDS tranche of the reference portfolio. Investors often want notes both listed and rated.

The second half of Chapter 12 covers SME securitization in Europe and the United States, secured loan trusts, and the Basel II regulatory capital framework. That is the next post.

Previous: CDO Managers: The Good, the Bad, and the Valued Few Next: SME Securitization and Basel II: Europe vs America