ABS and CLO Structured Credit
Asset-backed securities (ABS) and collateralized loan obligations (CLOs) transform heterogeneous collateral cash flows into tranches whose losses, coupons, and principal depend on a contractual waterfall. Their risk is governed by timing, correlation, trigger mechanics, and manager behavior as much as by average collateral default probability.
Waterfall before spread
Start by implementing legal cash-flow priority. A simplified waterfall collects collateral interest and principal, pays senior expenses and coupons, then allocates remaining funds by seniority. Defaults reduce collateral balance; recoveries arrive later; excess spread may absorb losses before principal writedowns.
interest_available = collateral_interest - fees - hedge_cost
pay senior interest -> mezzanine interest -> junior interest
principal_available = collections + recoveries + redirected interest
pay senior principal -> mezzanine principal -> residual
| Feature | Quantitative consequence | Naive error |
|---|---|---|
| sequential pay | senior deleverages faster | assuming pro rata principal |
| revolving period | reinvestment changes pool | freezing collateral |
| triggers | cash flow diverts senior | ignoring discontinuity |
| recovery lag | timing loss and discounting | booking recovery on default date |
| excess spread | first-loss protection | treating it as guaranteed |
ABS collateral commonly has amortization and prepayment; CLO collateral is usually floating-rate leveraged loans with reinvestment discretion. A model should separately represent contractual rules and assumptions about collateral evolution.
Tranche loss is nonlinear
Let portfolio loss fraction be L, tranche attachment A, and detachment D. The terminal tranche loss fraction is:
TL(L) = min(max(L - A, 0), D - A) / (D - A)
That formula is useful for intuition but insufficient for cash-flow valuation: it omits default timing, coupon diversion, recoveries, and trigger feedback. Correlation is especially important around attachment points. Two pools with identical expected loss can give very different mezzanine outcomes.
import numpy as np
def tranche_loss(portfolio_loss, attach, detach):
width = detach - attach
return np.clip(portfolio_loss - attach, 0.0, width) / width
def discounted_cashflows(cashflows, times, spread, base_dfs):
return np.sum(cashflows * base_dfs * np.exp(-spread * times))
Use simulation or scenario grids for collateral defaults, recoveries, prepayments, and spreads. Calibrate scenario severity to observable tranche prices where available; a point estimate of correlation should never be presented as a physical truth.
CLO-specific state variables
CLO overcollateralization (OC) and interest-coverage (IC) tests redirect cash to senior noteholders when ratios breach thresholds. They protect senior debt but can starve junior tranches and equity. The collateral manager may trade, reinvest principal, or exploit rating and spread constraints, so collateral composition is endogenous.
| Metric | Definition sketch | What it misses |
|---|---|---|
| OC ratio | adjusted collateral / senior notes | future trading and haircuts |
| WARF | weighted rating-factor proxy | default clustering |
| WAS | collateral spread average | funding and price volatility |
| WAL | average maturity | extension and reinvestment risk |
Model loan ratings, prices, maturity, industry, obligor, and recovery assumptions at the position level. Stress downgrade-driven haircuts and par erosion separately from realized defaults; market value can matter before cash losses appear.
Relative value and hedging
Compare tranche spreads after normalizing for expected collateral, structure, reinvestment period, manager quality, financing, and liquidity. A wider spread can be compensation for complexity or a trigger-cliff exposure rather than mispricing. Hedge broad credit beta with indices only as a partial hedge: index spread moves do not replicate idiosyncratic default, recovery, or waterfall convexity.
For single-name credit inputs, hazard-rate modelling should be internally consistent with CDS pricing. Stress correlated sector defaults, recovery dispersion, delayed collections, and collateral spread widening simultaneously rather than one variable at a time.
Key takeaways
- Encode the contractual waterfall before estimating spread or expected loss.
- Tranche value depends on timing and correlation, not merely average pool loss.
- CLO OC/IC triggers create discontinuous senior protection and junior downside.
- Collateral-manager behavior and reinvestment make static-pool assumptions unreliable.
- Relative value requires scenario analysis across defaults, recoveries, triggers, and liquidity.