Firmwide Risk Aggregation with Copulas and Capital Allocation

Book: Financial Risk Management: Applications in Market, Credit, Asset and Liability Management and Firmwide Risk Authors: Jimmy Skoglund & Wei Chen ISBN: 978-1-119-13551-7

Chapter 8 asks: what is total risk across market, credit, liquidity, and operational silos? Silo models are necessary but not sufficient. The 2007 crisis showed that.

Top-down vs bottom-up

Top-down (this chapter): compute risk in each silo independently, then aggregate with a correlation or copula model.

Bottom-up (Chapter 9): simulate joint scenarios across all risk factors, compute P&L per scenario, aggregate at the firm level.

In practice, banks mix both. Some books aggregate bottom-up and feed into a top-down layer for the rest.

Simple sum is not enough

Basel II regulatory capital is roughly additive across risk types. But summing VaR numbers is not conservative. Diversification can reduce true firmwide risk by 10-30%, but compounding effects can also make the sum understate risk.

Linear aggregation uses the covariance formula on standalone VaR or economic capital numbers. With correlation matrix rho_ij:

EC_firm = sqrt(sum of EC_i^2 + 2 * sum of rho_ij * EC_i * EC_j)

Zero correlation gives the Euclidean sum (diversified). Perfect correlation gives the arithmetic sum (worst case).

Copula-based aggregation

Linear models assume elliptical distributions. Copulas separate marginal loss distributions from the dependence structure. Gaussian copulas, t-copulas, and mixed/hierarchical copulas give more flexibility for tail dependence.

Hierarchical copulas are especially useful when some risks aggregate bottom-up first (e.g., trading book market and credit) and then join banking book risks at a second level.

Non-financial risks

Operational, strategic, and reputational risks can enter the same framework if you can produce a loss distribution and assign correlations. Expert judgment drives the inputs.

Capital allocation after aggregation

Once firmwide economic capital is computed, it must be allocated back to business lines. Euler allocation splits diversified capital into additive contributions that sum to the total.

This feeds RAROC, limit setting, and performance measurement. Without allocation, diversification benefits are invisible to business unit managers.

Concentration and diversification metrics also come from the decomposition.

Example

The book’s Table 8.1 shows five sub-risks (market, credit, operational, funding, other). Correlated aggregation gives lower total risk than simple addition. Each sub-risk gets a percentage contribution to the total.

Regulation

Basel Committee guidance on risk aggregation and firmwide capital recognizes that simple summation is a starting point, not the answer. ICAAP expects banks to justify diversification benefits with data and stress tests.

Mixed and hierarchical copulas

Mixed copulas let different risk pairs use different dependence structures. Market and credit might use a t-copula with tail dependence. Market and operational might use Gaussian. Hierarchical aggregation builds a tree: sub-portfolios first, then firm level.

This matches reality better than one correlation matrix for five unlike risks.

Diversification benefit reporting

Banks often report diversification benefit as the difference between summed standalone capital and aggregated firm capital. Board packs show how much capital is saved from diversification versus how much is consumed by concentration. Euler allocation splits the diversified total so each desk sees a fair capital charge.

Regulatory perspective

Basel Committee papers acknowledge that simple sum is conservative in some dimensions and optimistic in others. ICAAP must justify correlation assumptions with historical data and stress evidence. Regulators remain skeptical of large diversification credits without documentation.

My take

Copulas are not magic. A Gaussian copula with wrong correlations is worse than a simple sum because it gives false comfort. The value is flexibility and the ability to model tail dependence.

The hierarchical approach matches how banks are actually organized. Aggregate within the trading book, aggregate within the banking book, then combine.

Chapter 8 sets up Chapter 9 cleanly. Top-down aggregation gives a fast firmwide number. Bottom-up stress testing validates whether the correlations assumed here break under real scenarios.

Previous: FTP Profitability and Regulation Next: Firmwide Scenario and Stress Approaches