FRM Exam Part II · Integrated Risk Management

Risk Aggregation and Risk Integration Methods

Q: What is the difference between risk aggregation and risk integration?

Aggregation is combining separate risk measures into one total. Integration is broader. It includes making measures comparable and modelling how risk types interact.

Q: Why does the additive approach overstate risk?

It assumes perfect dependence, so every risk type hits its worst level together. That ignores diversification and gives an upper bound.

Q: Why use a copula instead of a correlation matrix?

A copula lets you fit any marginal distribution to each risk and model dependence separately, including tail dependence. A correlation matrix alone captures only linear association.

Q: Does diversification benefit hold in a crisis?

Often it shrinks. Correlations between risk types tend to rise in stress, so the aggregate moves closer to the simple sum.

Updated 11 October 2026 · Fact-checked

Risk aggregation combines market, credit, operational and other risk measures into one firm-wide figure. Methods are simple addition, a variance-covariance formula using correlations, and copulas that join full loss distributions. Addition ignores diversification. Correlation and copula methods capture it, but depend on assumptions that can fail in stress.

Understand Risk Aggregation and Risk Integration

A bank measures market, credit and operational risk separately, often with different models, horizons and confidence levels. Management and regulators still need one number for total capital. Risk aggregation is the step that combines these separate measures. Risk integration is the wider job of making the risk types comparable and modelling how they interact.

The simplest method is additive. You add the stand-alone capital figures. This assumes the risks are perfectly dependent and all hit at their worst together. It is conservative and easy, but it gives no credit for diversification. Diversification benefit is the gap between the sum of stand-alone risks and the aggregate risk.

The next method uses correlations. You treat each risk type's economic capital as a standard-deviation-like quantity and combine them with a correlation matrix. This is the variance-covariance method. It is quick and transparent. It works exactly only when losses are jointly normal (or elliptical), and a single correlation number cannot describe dependence in the tails.

A copula separates the marginal distribution of each risk from the dependence structure between them. You fit each risk's loss distribution, then link them with a copula, simulate joint losses and read off the VaR or expected shortfall of the total. A Gaussian copula has no tail dependence. A Student-t copula has symmetric tail dependence, so extreme losses cluster more. Choice of copula can change total capital sharply.

Pitfalls are common exam material. Risk measures differ in horizon (for example one-year versus ten-day) and confidence level, so they must be made consistent first. Correlations are hard to estimate, especially across risk types with little data. Correlations rise in crises, so diversification benefit shrinks when you need it. Also, VaR is not subadditive in general, so aggregated VaR can exceed the sum of parts for non-elliptical distributions.

Key formulas to remember

Additive aggregation: Total = C₁ + C₂ + ... + Cₙ; Equals the correlation-based result when all correlations are 1. Gives zero diversification benefit.
Correlation-based aggregation (two risks): Total = √(C₁² + C₂² + 2ρC₁C₂); C are stand-alone capital figures at the same horizon and confidence level. Exact for jointly normal losses.
General correlation-based aggregation: Total = √(Σᵢ Σⱼ ρᵢⱼ Cᵢ Cⱼ); ρᵢᵢ = 1. Uses the full correlation matrix.
Diversification benefit: Benefit = ΣCᵢ − Total; Benefit % = Benefit ÷ ΣCᵢ; Falls as correlation rises. It is zero at ρ = 1.
Sklar's theorem: F(x₁,...,xₙ) = C(F₁(x₁),...,Fₙ(xₙ)); Any joint distribution equals a copula C applied to its marginals.

How to solve Risk Aggregation and Risk Integration questions

Use this method for any aggregation question, numerical or conceptual.

1Identify the risk types and their stand-alone figures. Check each is at the same horizon and confidence level.
2Identify the method the question names or implies: additive, correlation matrix or copula.
3For additive, sum the figures. For correlation-based, apply the square-root formula with the given correlations.
4Compute diversification benefit as the sum of stand-alone figures minus the aggregate, and express it as a percentage if asked.
5For copula questions, decide which feature matters: marginals, dependence type or tail dependence (Gaussian has none, Student-t has some).
6Check the assumptions. Ask whether normality, stable correlations or equal horizons hold.
7Interpret the result: a lower total means a diversification credit, and say how it could vanish in stress.

Quickest way: Bounds and sanity check

When to use it: Use when options are numerically spread and time is short.

Compute the additive sum. This is the upper bound for the correlation method with ρ ≤ 1.
Compute the ρ = 0 result, the square root of the sum of squares. This is a lower reference when correlations are positive.
The true answer must lie between these two. Eliminate options outside the range.
For two risks, plug the given ρ into the formula once to pick the final option.
For concept questions, remember: higher correlation or tail dependence means less diversification.

Common mistakes in Risk Aggregation and Risk Integration

Adding capital figures and calling it the diversified total.
Addition feels natural and safe.
Fix: Addition assumes ρ = 1. Use the correlation formula when correlations are given.
Forgetting the square root in the correlation formula.
Students stop after summing squares and cross terms.
Fix: Always take √ at the end. The total must be below the simple sum when ρ < 1.
Aggregating figures at different horizons or confidence levels.
Market risk is often 10-day, credit and operational are one-year.
Fix: Scale to a common horizon and confidence level before combining.
Saying a Gaussian copula captures tail dependence.
Confusing correlation with tail behaviour.
Fix: The Gaussian copula has no tail dependence. Student-t does, and clusters extreme losses.
Assuming diversification benefit is stable.
Correlations are estimated in calm periods.
Fix: State that correlations tend to rise in stress, so the benefit can shrink exactly when needed.
Assuming VaR always shrinks on aggregation.
Mixing up normal-case results with the general rule.
Fix: VaR is not subadditive in general. The sum-bound holds for elliptical distributions.

Worked examples

Example 1

A bank has stand-alone economic capital of $400 million for market risk and $300 million for credit risk, at the same horizon and confidence level. The correlation is 0.5. Find aggregate capital and the diversification benefit.

Show the solution

Total = √(400² + 300² + 2 × 0.5 × 400 × 300).
400² = 160,000 and 300² = 90,000.
Cross term = 2 × 0.5 × 120,000 = 120,000.
Sum = 160,000 + 90,000 + 120,000 = 370,000.
√370,000 ≈ 608.3.
Additive sum = 700. Benefit = 700 − 608.3 = 91.7.

Answer: Aggregate capital ≈ $608 million; diversification benefit ≈ $92 million (about 13%).

Example 2

A risk team replaces a Gaussian copula with a Student-t copula, keeping the same marginal loss distributions and the same correlation parameter. What is the likely effect on aggregate tail risk, and why?

Show the solution

Marginals are unchanged, so only the dependence structure changes.
The Gaussian copula has no tail dependence. Extreme losses in different risks are nearly independent far out in the tail.
The Student-t copula has positive tail dependence, so extreme losses are more likely to occur together.
Joint extreme losses become more likely, which raises high-confidence VaR and expected shortfall of the total.

Answer: Aggregate tail risk rises and diversification benefit falls, because the Student-t copula adds tail dependence.

Exam tips

Check units and confidence levels first. Questions often hide a mismatch.
Memorise the two-risk formula and the ρ = 1 and ρ = 0 special cases for quick bounds.
Know the copula vocabulary: marginals, dependence, tail dependence, Gaussian versus Student-t.
Expect interpretation questions: why diversification benefit may be overstated and what stress does to it.

Practice questions from Integrated Risk Management

Risk Aggregation and Risk Integration: frequently asked questions

What is the difference between risk aggregation and risk integration?

Aggregation is combining separate risk measures into one total. Integration is broader. It includes making measures comparable and modelling how risk types interact.

Why does the additive approach overstate risk?

It assumes perfect dependence, so every risk type hits its worst level together. That ignores diversification and gives an upper bound.

Why use a copula instead of a correlation matrix?

A copula lets you fit any marginal distribution to each risk and model dependence separately, including tail dependence. A correlation matrix alone captures only linear association.

Does diversification benefit hold in a crisis?

Often it shrinks. Correlations between risk types tend to rise in stress, so the aggregate moves closer to the simple sum.