A CDO divides a debt portfolio into credit-risk layers. Trace a principal loss through first-loss, mezzanine and senior tranches, and distinguish cash from synthetic exposure.
Collateralised debt obligation
A collateralised debt obligation (CDO) is a structured credit transaction that divides the credit risk of a debt portfolio into tranches with different loss priorities. Buying a tranche means taking a particular layer of the pool's risk: two investors exposed to the same borrowers can suffer very different percentage losses.
The portfolio can contain loans, bonds or structured securities, including tranches of other securitisations. A collateralised loan obligation specialises in loans. A CDO backed by other CDO tranches adds another layer of structuring, often called a CDO-squared. The label does not tell you which assets are inside, how concentrated they are or how the payments work.
What is actually being pooled?
In a cash CDO, a vehicle purchases debt assets and issues claims backed by them. Interest, repayments and asset-sale proceeds support payments to investors under the transaction documents. This applies securitisation to a credit portfolio; the assets need not all be corporate bonds.
In a synthetic CDO, credit derivatives such as credit default swaps create exposure to a reference portfolio. The structure need not buy the referenced debt. A funded tranche investor can lose contributed principal; an unfunded tranche investor promises contingent protection payments. Both bear tranche credit risk. Funded versus unfunded describes how that exposure is financed, rather than whether the underlying debt is owned. Gibson's Federal Reserve paper, §§1–2 explains these distinctions; the Basel Framework separately defines traditional and synthetic securitisations in CRE40.2–40.3.
Follow a loss through the structure
Consider a fictional cash CDO with USD 100 million of initial pool principal. Stipulate three layers: first-loss capital of USD 10 million, mezzanine principal of USD 20 million and senior principal of USD 70 million. Allocate realised principal credit losses strictly in that order. Ignore amortisation, fees, interest, reserves, excess spread and contractual triggers.
Suppose the pool incurs USD 15 million of credit loss after recoveries. The allocation is:
Layer
Initial principal, USD m
Principal lost, USD m
Principal remaining, USD m
Loss as % of own initial principal
First-loss
10
10
0
100%
Mezzanine
20
5
15
25%
Senior
70
0
70
0%
Total
100
15
85
15% of the pool
Fictional USD 100 million pool with USD 15 million credit loss: first-loss capital loses 10 of 10, mezzanine loses 5 of 20, and senior loses 0 of 70.
Strict principal-loss priority in a fictional structure. The mezzanine investor loses 25% of its capital while the pool loses 15%; the senior layer has no principal loss in this scenario.
For a layer that attaches after A million of pool losses and has width W million, its principal loss at pool loss L is
ℓ(L)=min(max(L−A,0),W).
All quantities here use the same USD-million basis. The mezzanine layer has A=10 and W=20, so L=15 gives ℓ=5. The senior layer attaches at A=30: it starts losing principal only when pool credit losses exceed USD 30 million. Default counts alone cannot establish whether that boundary is crossed; exposure sizes and recoveries matter.
The USD 85 million remaining is an accounting balance in this example, not a market valuation or a promise of future repayment. Nor is this loss ledger the complete cashflow waterfall. Real documents determine payment dates, coupon priorities, fees, reinvestment rules and tests that can divert cash. A tranche can fall in market value before it suffers a principal writedown.
The toy allocation conserves the loss: USD 10 million plus USD 5 million plus zero still equals USD 15 million. Subordination changes who bears it; it does not erase the borrowers' losses.
For a risk model, the difficult step comes before allocation: how likely are different pool losses, and when do they occur? Individual default probabilities do not specify the joint outcome. Default dependence, shared sector or geographic exposures, and uncertain recoveries can alter the chance of losses reaching a senior layer. The resulting tranche sensitivities depend on the portfolio, attachment points and model; there is no universal rule that a change in correlation moves every tranche the same way.
A useful analysis therefore combines collateral due diligence, the actual payment rules and stress tests of the loss model. A credit rating or a large subordinate cushion alone does not supply a price, a default forecast or a full account of liquidity and model risk.
Sources
Frank J. Fabozzi, Bond Markets, Analysis, and Strategies, 9th edition, Pearson, 2016, Chapter 15, pp. 335–337: collateral types, CDO structure and distribution rules.
Michael S. Gibson, Understanding the Risk of Synthetic CDOs, FEDS 2004-36, §§1–2 and 4–6: cash versus synthetic structures, funded versus unfunded exposure, tranche loss and risk sensitivity. Historical research; the example above uses its own stipulated amounts.
Basel Committee on Banking Supervision, Basel Framework CRE40, especially 40.2–40.6: traditional, synthetic and resecuritisation definitions. Current chapter checked 27 October 2026.
