FRM Exam Part II · Financial Correlation Modeling - Bottom-Up Approaches

Correlation Risk in CDOs and Tranche Sensitivity

Updated 11 October 2026 · Fact-checked

Correlation risk in CDOs is the change in tranche value when default correlation changes. Higher correlation raises equity tranche value and lowers senior tranche value. Mezzanine tranches can move either way. To solve questions, identify where the tranche sits, then ask whether higher correlation fattens the tails of the loss distribution.

Understand Correlation Risk in Structured Products and CDOs

A CDO pools credits and splits the pool's losses into tranches. The equity tranche absorbs the first losses. The mezzanine tranche absorbs the next losses. The senior tranche is hit only after the lower tranches are wiped out. Each tranche is defined by an attachment point and a detachment point.

Default correlation controls the shape of the portfolio loss distribution, not its mean. The expected pool loss stays the same whatever the correlation. What changes is how losses are spread. With low correlation, defaults are scattered and the loss is close to the average. With high correlation, defaults cluster: either very few names default, or many default together.

This is why tranches react differently. The equity tranche is hit by the first small losses. Higher correlation raises the chance of zero or very few defaults, so its expected loss falls and its value rises. The equity tranche is therefore long correlation. The senior tranche is hit only in extreme clustering. Higher correlation fattens that tail, so its expected loss rises and its value falls. The senior tranche is short correlation.

Mezzanine tranches sit in the middle. Their sensitivity is mixed and depends on attachment points, pool spread and maturity. Often it is small, and it can change sign. Because the pool's expected loss is fixed, a fall in expected loss on some tranches must be offset by a rise on others (weighted by tranche size).

In practice the market quotes tranches using the one-factor Gaussian copula. The correlation that makes the model match a tranche's market price is its implied correlation. Implied correlations differ across tranches, which is the correlation smile (or skew). It shows that the single-correlation Gaussian copula is not a true description of the market. Base correlation fixes the quoting by treating each tranche as the difference of two equity tranches, each with its own correlation. Base correlation typically rises with detachment point, giving a skew.

Key formulas to remember

Tranche width: Tranche width = Detachment point − Attachment point; Example: a 3%-7% tranche has width 4%. Thin tranches are more sensitive to the loss distribution.
Tranche loss: Tranche loss = min(max(L − A, 0), D − A), where L is pool loss, A attachment, D detachment; Loss is zero below A and capped at D − A. Express as a percentage of tranche width by dividing by (D − A).
Expected pool loss: E[L] = PD × (1 − R), as a fraction of pool notional; Assumes homogeneous names: equal notionals and a common PD and recovery rate R. Independent of correlation. Correlation changes only the distribution around this mean.
Correlation sensitivity rules: Equity: value ↑ as ρ ↑. Senior: value ↓ as ρ ↑. Mezzanine: mixed.; Equity is long correlation; senior is short correlation.
Base correlation decomposition: Tranche [A, D] = Equity(0, D) − Equity(0, A); Each equity piece is priced with its own base correlation, taken from the calibrated base correlation curve.
Gaussian copula one-factor model: Ai = √ρ × M + √(1 − ρ) × Zi; M is the common factor, Zi the idiosyncratic factor. Name i defaults if Ai falls below a threshold set by its PD.

How to solve Correlation Risk in Structured Products and CDOs questions

Use this approach for any question on correlation and tranche values.

1Identify the tranche from its attachment and detachment points: equity (lowest), mezzanine (middle), senior or super senior (top).
2State what changes: correlation moves, while the pool's expected loss stays fixed.
3Translate higher correlation into the loss distribution: more probability on very low and very high pool losses, less on middle losses.
4Ask which losses hit the tranche. Equity is hit by small losses, so its expected loss falls with higher correlation. Senior is hit by extreme losses, so its expected loss rises.
5Convert expected loss to value or spread: lower expected loss means higher value and a lower fair spread; higher expected loss means the opposite.
6For smile or base correlation questions, recall that implied correlation differs by tranche, and base correlation usually rises with detachment point.
7Check your direction against the rule: equity long correlation, senior short correlation, mezzanine ambiguous.
8Pick the option that matches both the direction and the reason given.

Quickest way: Long or short correlation in ten seconds

When to use it: Use for direction-of-change questions under time pressure.

Locate the tranche: bottom means long correlation, top means short correlation.
Higher correlation: equity value up, senior value down.
Equity spread falls when correlation rises; senior spread rises.
If the option says mezzanine always rises or always falls, be suspicious.
For smile questions, remember that one Gaussian correlation cannot fit all tranches.

Common mistakes in Correlation Risk in Structured Products and CDOs

Saying higher correlation raises the expected loss of the whole pool.
Students link correlation with more risk in general.
Fix: Expected pool loss depends on PD and recovery only. Correlation redistributes loss between tranches.
Treating the equity tranche as short correlation because it is the riskiest.
Risky is confused with sensitive in the wrong direction.
Fix: Equity gains when defaults cluster, because more scenarios have few or no defaults. Equity is long correlation.
Claiming the mezzanine tranche always behaves like equity or like senior.
Students want a clean rule for all three tranches.
Fix: Mezzanine sensitivity is mixed and depends on attachment points, spreads and maturity. Say it can change sign.
Confusing a correlation smile with a volatility smile.
Both use the word smile.
Fix: The correlation smile is implied correlation plotted against tranche detachment or strike. It shows the Gaussian copula is misspecified.
Thinking base correlation is the same as the tranche's implied (compound) correlation.
Both are backed out of the Gaussian copula.
Fix: Base correlation prices equity tranches from zero to each detachment point. Compound correlation prices each tranche on its own.
Forgetting that thin senior tranches still lose value when correlation rises.
Senior tranches are seen as safe.
Fix: Senior tranches carry tail risk. Higher correlation raises the probability of the extreme losses that reach them.

Worked examples

Example 1

A bank is long the 0%-3% equity tranche and short protection on the 12%-100% senior tranche of the same CDO. Market correlation rises. Holding spreads and recovery fixed, what is the likely effect on each position's value and on the combined position?

Show the solution

The pool's expected loss is unchanged, since PD and recovery are unchanged.
Higher correlation clusters defaults: more weight on very low and very high pool losses.
Being long the equity tranche means the bank holds its credit risk, which is the same as having sold protection on it. Equity is hit by small losses, so its expected loss falls and its value rises. The bank gains on this position.
Being short protection on the senior tranche means the bank has sold protection and bears the senior tranche's credit risk. Senior is hit only by extreme losses, so its expected loss rises. The bank loses on this position.
The bank is exposed to credit risk on both tranches, but the correlation effects run in opposite directions: a gain on equity and a loss on senior. The net result depends on the notionals and the correlation sensitivities of each position.

Answer: The equity position gains and the senior short-protection position loses. The bank takes credit risk on both tranches, so the net effect depends on position sizes and sensitivities. Holding both does not make the bank hedged against correlation.

Example 2

In a CDO the 0%-3% tranche has base correlation 20% and the 0%-7% tranche has base correlation 30%. Describe how the 3%-7% tranche is valued using base correlation, and what the pattern of base correlations indicates.

Show the solution

Write the tranche as a difference: 3%-7% tranche = Equity(0%-7%) − Equity(0%-3%).
Value the 0%-7% equity piece using the 30% base correlation.
Value the 0%-3% equity piece using the 20% base correlation.
Subtract the second value from the first to get the 3%-7% tranche value.
The base correlations differ and rise with detachment point (20% to 30%). This is the typical pattern and gives an upward base correlation skew.

Answer: The 3%-7% tranche equals Equity(0-7%) at 30% minus Equity(0-3%) at 20%. The upward skew, which is typical, shows that one flat Gaussian copula correlation does not fit the market prices of all tranches.

Exam tips

Memorise: equity long correlation, senior short correlation, mezzanine mixed. Most direction questions reduce to this.
Always separate the pool's expected loss (unchanged) from the loss distribution (changed by correlation).
For smile questions, state that a non-flat implied correlation shows Gaussian copula misspecification.
In hedging cases, check whether the positions are long or short protection before deciding the sign of the gain.
Watch the wording: tranche value and tranche spread move in opposite directions.

Practice questions from Financial Correlation Modeling - Bottom-Up Approaches

Correlation Risk in Structured Products and CDOs: frequently asked questions

Why is the equity tranche long correlation?

Higher correlation makes defaults cluster, so scenarios with few or no defaults become more likely. The equity tranche is hit first by small losses, so its expected loss falls and its value rises.

Why does the senior tranche lose value when correlation rises?

Senior tranches are hit only when many names default together. Higher correlation fattens that tail, raising the senior tranche's expected loss and lowering its value.

What is the correlation smile in CDOs?

It is the pattern of implied correlations that differ across tranches when each is priced with the Gaussian copula. If the model were right, one correlation would fit all tranches. The pattern shows it does not.

How is base correlation different from implied correlation?

Base correlation prices equity tranches from zero up to each detachment point, then gets other tranches by subtraction. Implied (compound) correlation prices each tranche separately. Base correlation is better behaved and gives a clearer skew.