What is a bespoke CDO (bespoke tranche)?
Definition
– Bespoke CDO (also called a bespoke tranche or bespoke tranche opportunity — BTO): a custom-constructed slice of a collateralized debt obligation (CDO) that a dealer builds to match a particular investor’s desired risk and return profile. Investors in a bespoke tranche typically buy a single tranche; the dealer keeps or hedges the rest.
– Collateralized debt obligation (CDO): a structured product that repackages a pool of credit exposures (loans, bonds, or credit derivatives) into layers, called tranches, each with different risk/return characteristics.
– Tranche: one of the ordered layers of a pooled debt structure. Losses are allocated from the most junior tranche upward; higher-risk tranches pay higher expected yields.
– Synthetic CDO: a CDO that gains exposure via credit derivatives (usually credit default swaps, or CDS) rather than holding the underlying loans or bonds. Bespoke CDOs are often synthetic, which makes them highly customizable.
– Over‑the‑counter (OTC): traded directly between parties rather than on an exchange; OTC products typically have less transparency and liquidity.
How a bespoke tranche works (simple logic)
1. Dealer assembles a portfolio of credit exposures (often CDS on many reference entities).
2. That portfolio is sliced into tranches with specified attachment and detachment points (the loss thresholds that determine when a tranche suffers principal losses).
3. An investor purchases a single tranche and receives periodic payments (higher for riskier tranches). The dealer retains or hedges the remaining tranches.
4. If portfolio losses occur, the junior tranches absorb losses first. A buyer of a mezzanine or equity tranche will incur losses sooner than a senior tranche buyer.
Key features (rephrased)
– Customizable to specific credit exposures and attachment/detachment points.
– Potentially higher yields than comparable plain‑vanilla credit instruments.
– Often constructed from many reference names, so some diversification exists.
– Traded OTC, which tends to reduce transparency and secondary-market liquidity.
– Valuation depends on complex pricing models and assumptions about default probabilities and correlations.
– Frequently not rated by major agencies; pricing is typically established by the dealer and market participants.
Why bespoke tranches fell out of favor — and why they returned
– During the 2007–2009 financial crisis, highly structured tranches and synthetic CDOs played a central role in amplifying losses and spreading them across markets. Complexity, model risk, and illiquidity contributed to the systemic effects.
– After the crisis, issuance declined sharply. From the mid‑2010s some dealers and investors returned to bespoke tranches (sometimes labeled BTOs), with greater attention to model governance and due diligence. Large dealers like Citigroup have been active in offering standardized portfolio views and publishing tranche prices to improve transparency.
Pros and cons — practical summary
Pros
– Tailored exposure: pick specific names or sectors and choose attachment/detachment points that match a desired payoff.
– Higher income: yields can exceed those on plain‑vanilla corporate bonds if you accept greater credit and liquidity risk.
– Risk transfer: allows originators or banks to move credit risk to willing counterparties.
Cons
– Illiquidity: small or nonexistent secondary market for highly customized tranches.
– Model and counterparty risk: valuations depend on complex models (default probabilities, correlation assumptions) and on the dealer’s solvency.
– Limited transparency: OTC structure and bespoke design reduce comparability and market pricing signals.
– Usually unsuitable for most retail investors due to complexity and minimum size; typically used by institutions.
Due-diligence checklist (short)
– Know the structure: identify attachment/detachment points and exactly which losses affect your tranche.
– Review reference portfolio: number of names, concentration, sectors, average credit quality.
– Ask about modeling: which models and correlation assumptions were used, and how sensitive the tranche price is to those assumptions?
– Counterparty and dealer risk: who is on the other side? What are their hedges and creditworthiness?
– Liquidity plan: how and when could you exit the position? Is there any secondary market?
– Legal docs and triggers: read governing agreements, default triggers, and payment waterfall mechanics.
– Independent valuation: obtain or run independent pricing models; challenge dealer quotes.
– Regulatory and tax treatment: check whether the instrument’s status under applicable rules affects capital, margin, or tax.
Small numeric example (worked)
Assume a reference portfolio with notional $100 million is tranched as:
– Equity tranche: 0%–3% (absorbs first $3M of losses)
– Mezzanine tranche: 3%–10% (next $7M of losses)
– Senior tranche: 10%–100% (remaining $90M)
An investor buys the mezzanine tranche (notional = $7M) and receives a higher coupon than senior debt.
Scenario: Actual portfolio losses = 5% of $100M = $5M
Loss allocation:
1. Equity tranche absorbs first $3M → equity wiped out.
2. Remaining $2M of loss hits mezzanine tranche.
Investor outcome:
– Mezzanine tranche notional = $7M. Loss = $2M → remaining principal = $5M (about 28.6% loss on the mezzanine position).
– Senior tranche unaffected.
This example shows how tranche position determines loss sensitivity: a relatively modest 5% portfolio loss produces a disproportionately large percentage loss for the mezzanine holder.
Important assumptions and model risk
– The above ignores recovery rates, default timing, and correlation structure,
all of which materially affect tranche losses.
Recovery rates and loss given default (LGD)
– Definitions: Probability of default (PD) is the chance a borrower defaults. Loss given default (LGD) is the fraction of exposure lost when default occurs (1 − recovery rate). Exposure at default (EAD) is the outstanding exposure at default.
– Core formula (single obligor): Expected loss (EL) = PD × LGD × EAD. For a portfolio, EL is the sum of ELs across obligors, but tranche losses depend on how those losses aggregate and where they fall relative to tranche attachment points.
– Practical consequence: Two portfolios with the same PD can produce very different tranche outcomes if LGDs differ. Higher LGDs shift more loss into mezzanine and senior tranches.
Timing and default sequencing
– Defaults occur over time, not all at once. Early defaults can trigger triggers in deal docs (e.g., early amortization, liquidity draws) that change cashflow priorities.
– Reinvestment features and amortization schedules change tranche exposure over time. Always check the transaction waterfall and triggers.
Correlation and concentration risk
– Correlation (how likely defaults move together) is the key driver of tranche risk. High correlation increases the probability of large, simultaneous losses that pierce mezzanine or senior tranches.
– Bespoke CDOs often concentrate specific sectors or exposures. Concentr
Concentration — holding many exposures in the same industry, geography, or to a few large obligors — magnifies idiosyncratic correlation and makes tranche outcomes far less predictable than a broadly diversified pool. In practice, a handful of large losses in a concentrated bespoke portfolio can pierce mezzanine or senior tranches that would be safe in a broadly diversified index-style CDO.
Valuation and model risk
– Structural complexity: Valuation depends on inputs that are often proprietary or unverifiable (single-name PDs, pairwise correlations, recovery/LGD assumptions, default timing). Small changes in key inputs can move tranche values a lot.
– Model risk: Gaussian copula, one-factor models, and other common approaches impose specific dependence structures that may not reflect reality. Always treat model outputs as conditional on assumptions.
– Market illiquidity: Bespoke tranches are thinly traded. Mark-to-market prices may be stale or based on dealer quotes rather than actual trades.
– Document sensitivity: Cashflow waterfalls, trigger levels, definitions of default, and re-investment mechanics change valuation materially. Two deals with identical portfolio lists can value differently if their legal docs differ.
Operational, counterparty, and legal risks
– Documentation: Ambiguous definitions (what counts as default, timing, cure periods) create legal risk. Review prospectus/indenture and any supplemental agreements.
– Servicer/issuer risk: Performance depends on the servicer’s accuracy in reporting, loss collection, and timing. Counterparty credit risk (e.g., swap counterparties funding coupons) matters.
– Reporting and transparency: Bespoke deals often disclose less, so monitoring requires active diligence.
Practical due-diligence checklist (step-by-step)
1. Confirm deal basics
– Notional, tranche attachment A and detachment D (expressed as % of portfolio notional).
– Waterfall and priority of payments.
– Reinvestment period, amortization rules, early amortization triggers.
2. Asset-level review
– Get the loan/bond list: obligor names, exposures, vintages, original balances, remaining balances.
– Obtain PD and LGD inputs and source (internal models, ratings, market-implied).
3. Legal/document review
– Read definition sections for “default”, “credit event”, “related party”, “interest shortfall”.
– Identify who controls remedial actions and note consent/transfer restrictions.
4. Modeling and stress testing
– Re-run valuations under alternative PD, LGD, and correlation assumptions.
– Conduct scenario analysis: single-name default(s), sector shock, systemic stress.
5. Operational checks
– Confirm trustee/servicer identity, counterparty credit support, and reporting cadence.
– Ask for historical performance files if similar vintage deals exist.
6. Liquidity and exit
– Check trade history and bid/offer spreads for comparable tranches.
– Plan exit scenarios (hold to maturity, sell, hedge).
Key formulas and a worked numeric example
– Portfolio expected loss (simplest, unconditional):
Expected loss (%) = PD * LGD
where PD = probability of default (fraction) and LGD = loss given default (fraction).
– Tranche loss mapping (static, simple rule):
Let L_portfolio = portfolio cumulative loss (%) then tranche cumulative loss (%) =
min(max(L_portfolio − A, 0), D − A)
where A = attachment point (%) and D = detachment point (%).
Worked example
Assumptions:
– Portfolio notional = 100 (expressed in % terms: 100%).
– Obligor-level PD (annual) = 2% -> roughly 0.02.
– LGD (given default) = 40% -> 0.40.
– Equity tranche: A = 0%, D = 5% (absorbs first 0–5% of losses).
– Mezzanine tranche: A = 5%, D = 15% (absorbs 5–15%).
Step 1 — expected portfolio loss (annual, simple)
– Expected loss = PD * LGD = 0.02 * 0.40 = 0.008 = 0.8% of portfolio notional.
Interpretation: On average, 0.8% of portfolio notional losses per year — which would be fully absorbed by the equity tranche in this example.
Step 2 — single stress scenario
– Suppose an adverse shock causes 10% of obligors to default in a short period.
– Portfolio loss = 10% * LGD (0.40) = 4% loss.
– Equity tranche (0
– Equity tranche (0–5%) takes the full 4% loss; mezzanine and senior are unaffected.
Step 3 — larger stress scenarios (worked examples)
– Scenario A — 30% defaults in a short period
– Portfolio loss = 30% * LGD (0.40) = 12% of portfolio.
– Loss allocation by tranche:
– Equity (0–5%): absorbs first 5% → equity loss = 5% (fully wiped out).
– Mezzanine (5–15%): next 10% capacity → absorbs next 7% of losses (12% − 5% = 7%) → mezzanine loss = 7% (70% of mezzanine notional).
– Senior (15%+): remaining losses = 0% (12% < 15% so senior untouched).
– Tranche loss rates (loss as a share of each tranche’s notional):
– Equity tranche: 5% / 5% = 100% loss.
– Mezzanine tranche: 7% / 10% = 70% loss.
– Senior tranche: 0% / 85% = 0% loss (senior notional here = 85% of portfolio).
– Scenario B — extreme stress: 50% defaults
– Portfolio loss = 50% * 0.40 = 20% of portfolio.
– Allocation:
– Equity: 5% (fully consumed).
– Mezzanine: 10% (fully consumed; 5–15%).
– Senior: remaining = 20% − 15% = 5% of portfolio absorbed by senior.
– Tranche loss rates:
– Equity: 5/5 = 100%.
– Mezzanine: 10/10 = 100%.
– Senior: 5% / 85% ≈ 5.88%.
Step 4 — expected (baseline) tranche losses (how to compute)
Use the baseline expected portfolio loss (PD * LGD). For a tranche with attachment A and detachment D (both expressed as portfolio percentage points):
– Absolute expected tranche loss (as percent of portfolio) = clamp(EPL − A, 0, D − A),
where EPL = expected portfolio loss (PD * LGD) and clamp(x, low, high) = min(max(x, low), high).
– Tranche expected loss rate (fraction of tranche notional) = (Absolute expected tranche loss) / (D − A).
Worked baseline example using previously computed EPL = 0.8%:
– Equity (0–5%): expected absolute loss = clamp(0.8% − 0%, 0, 5%) = 0.8% → expected loss rate = 0.8% / 5% = 16% of equity notional.
– Mezzanine (5–15%): expected absolute loss = clamp(0.8% − 5%, 0, 10%) = 0 → expected loss rate = 0%.
– Senior: expected absolute loss = 0 → expected loss rate = 0%.
Practical checklist for tranche loss analysis
1. Inputs to gather
– PD: marginal/default probability (per period).
– LGD: loss given default (fraction of exposure lost when default occurs).
– Tranche attachment (A) and detachment (D).
– Portfolio concentration and single-name weights.
– Correlation assumptions (between obligors’ defaults).
– Timing assumptions (all defaults instant vs distributed over time).
2. Scenario analysis
– Run simple deterministic shocks (e.g., x%
% continued %
2. Scenario analysis
– Run simple deterministic shocks (e.g., x% portfolio loss or y% of names defaulting) and compute tranche impairment using the clamp function: clamp(z,0,thickness) = min(max(z,0), thickness). This gives a quick lower-bound on expected tranche loss under a given portfolio loss level.
– Vary key drivers one at a time (PD, LGD, correlation, concentration) to build sensitivity tables
– Continue sensitivity testing into joint moves. Example grid: PD ∈ {0.5%,1%,2%}, LGD ∈ {40%,60%,80%}, rho ∈ {0.05,0.2,0.4}, concentration index ∈ {equal weights, top-10% weight = 30%}. For each cell compute tranche expected loss (EL) and the change relative to base case; present results as heat maps or tornado charts.
3. Simulation workstream — Monte Carlo (practical recipe)
– Purpose: produce a numerical distribution of portfolio losses consistent with your dependence model (copula, factor model) and heterogeneity (different PDs, LGDs, weights).
– Inputs: list of N obligors with exposure weight wi (sum to 1), PDi (per horizon), LGDi, correlation structure (e.g., single-factor rho_i or full correlation matrix), number of draws S (e.g., 50k–200k), recovery timing and discount