Asset-Backed Security

Securitisation is finance's packaging machine: take a pool of loans that individually could never trade — car loans, credit-card balances, student loans, equipment leases, even music royalties — put them in a legal box, and sell bonds backed by the box's cash flows.

The box's bonds come in slices ("tranches") of different safety. Losses from defaulting borrowers hit the bottom slice first; only when it's wiped out do losses touch the next one up. The top ("senior") tranche is thus insulated by everything below it and earns the lowest yield; the bottom ("equity") absorbs first losses for the fattest potential return.

For the original lender, securitisation converts illiquid loans into fresh cash to lend again. For investors, it manufactures the exact risk level they want from raw material that had no market of its own.

The structure

• SPV: a bankruptcy-remote vehicle buys the assets — investors are exposed to the pool, not the originator ("true sale").
• Waterfall: contractual rules routing interest and principal top-down, and losses bottom-up.
• Credit enhancement: subordination, excess spread (pool interest > bond coupons), overcollateralisation, reserve accounts.
• Triggers: performance covenants that redirect cash to seniors if the pool deteriorates.

What can go wrong

2008 taught the failure modes: correlated collateral (all tranches die together if losses are systemic), model overconfidence in ratings, originators keeping no "skin in the game" (now mandated: 5% risk retention in EU/US), and complexity opacity (CDO-squared). Modern consumer ABS — simpler, shorter, amortising — sailed through recent stresses.

Reading a deal

Key metrics: collateral type and seasoning, WAL (weighted average life), subordination %, excess spread, delinquency/loss curves vs. base case, and originator quality. Ratings are an opinion on the tranche, not the pool.

Tranche mathematics

A tranche with attachment \(a\) and detachment \(d\) on pool loss \(L\) pays losses \(\min(\max(L-a,0), d-a)\) — a call-spread on pool losses. Its expected loss:

Everything hinges on the loss distribution \(\mathbb{P}(L > x)\), which depends critically on default correlation: higher correlation fattens both tails — seniors get riskier, equity actually safer (more scenarios with zero losses).

Modelling approaches

• Consumer pools: actuarial — project default/prepay/severity vectors from vintage curves, run the waterfall engine, discount tranche flows at spread benchmarks; stress scenarios define ratings.
• Correlation-sensitive structures: factor models (Gaussian copula one-factor, \(\rho\) calibrated to tranche markets) — with the well-known caveat that copula correlation is a quoting device, not physics.

Valuation outputs

Tranches quote at discount margin / spread to WAL; analytics report credit-adjusted WAL, break-even CDR (constant default rate the tranche survives), and multiple-of-base-case-loss coverage — the structurer's language of protection.