Collateral Risk Management: Running the Collateral Like a Risk Asset

The most common failure in collateralized finance is not that collateral had no value — it is that institutions treated collateral as a fixed "yes/no" box rather than as a living portfolio of risk that must be actively managed. This page presents the discipline as professional collateral desks and bank risk functions practice it: what risks exist in a collateral pool, how each is measured and controlled, and what the operating systems behind them look like.

The Risk Inventory: Six Risks in Every Collateral Pool

1. Market risk — the collateral's value falls before it can be liquidated. Managed through haircuts, revaluation frequency, and call triggers (see valuation).
2. Concentration risk — value is piled into one asset, issuer, sector, or geography, so a single shock can wipe out a large share of the pool. Managed through limits and diversification floors.
3. Correlation (self-collateralization) risk — the collateral falls because of the same event that triggers the underlying default. Posting a bank's own debt to secure credit to that bank is the textbook example: the hedge is zero in the exact scenario it is supposed to cover.
4. Liquidity risk — the asset cannot be sold in the required size and time, regardless of price. Managed through liquidity tiers, forced-sale horizon testing, and exclusions of thin assets.
5. Custody and operational risk — loss of the asset itself: failed transfers, unregistered pledges, missing certificates, fraud, or custodian failure. Managed through custody standards, tri-party agents, and reconciliation.
6. Legal and jurisdictional risk — the enforcement right fails in the jurisdiction where the asset sits, or cross-border transfer is restricted by capital controls. Managed through netting and collateral agreements (CSA/GMRA-style documentation), local-law opinions, and jurisdictional limits.

Loss Given Default: Measuring What Collateral Actually Saves

The performance of a collateral program is ultimately measured in loss given default (LGD): the share of the exposure lost when a borrower fails. For unsecured corporate lending, long-run average LGD in the US hovers around 55–60%; for well-structured secured lending it can fall to 15–30% or less — the difference being the collateral. Risk teams estimate LGD by asset class, borrower type, and economic cycle, because LGD is counter-cyclical: it is worst precisely in downturns when collateral prices are falling and sale horizons lengthen. Capital models under Basel reflect this: the internal-ratings approach requires banks to model the distribution of LGD, not a single point, and the standardized approach simply applies risk-weight steps by security type and LTV. A collateral desk that reports "95% of the book is secured" without an LGD decomposition is reporting a slogan, not a risk position.

Concentration Limits and Diversification

Concentration is the risk most often ignored until it is the story. Standard limit architecture in a collateral pool:

• Single-name limit — e.g., no more than 5–10% of pool value in one issuer or security (tighter for equities, since a single share can be 100% correlated with a single failure).
• Sector / asset-class limits — e.g., real estate at no more than X% of the bank's collateralized book; high-yield bonds capped separately from IG.
• Geographic limits — both for market risk (one housing market) and legal risk (one jurisdiction's enforcement regime).
• Diversification benefits, quantified — not assumed — a properly modeled pool enjoys a real reduction in aggregate volatility from holding many low-correlation assets; risk systems compute the credit and mark the haircut schedule against the modeled, diversified pool rather than against each asset in isolation. The 2008 error was assuming diversification in a structure where default correlations were near one; the fix was to make the correlation input a tested, stressed assumption.

Monitoring: The Operating Rhythm of a Collateral Desk

Collateral management is continuous, with a standard operational rhythm:

• Daily: mark the entire pool at an independent source; recompute coverage ratios per counterparty and per pool; issue and chase margin calls and top-ups; reconcile positions with the custodian or tri-party agent; log valuation disputes.
• Weekly: limit utilization review (names, sectors, jurisdictions); liquidity ladder — how much of the pool can be liquidated in 1, 5, and 30 days, and at what modeled cost; haircuts review against realized volatility.
• Quarterly: stress testing (see below); LGD back-testing against any realized defaults; counterparty-level review of collateral quality mix; legal review of documentation for new counterparties and jurisdictions.

The technology behind this — collateral management systems that track every pledge, every valuation, every call, and every limit in real time — is itself a critical control. Much of the operational fragility exposed in 2008 (and in subsequent repo market squeezes) was systems and process fragility: mismatched data, manual reconciliations, and calls that could not be executed fast enough.

Stress Testing the Collateral Pool

A collateral pool is stress-tested the same way a trading book is, with scenarios matched to its composition: a 40% equity market drawdown with a 3× volatility regime; a 300bp credit spread widening across the corporate bond holdings; a 25% property price correction; a sovereign-currency event for the jurisdiction where enforcement occurs; and a liquidity scenario — forced liquidation of the pool within 10 business days, with haircuts widened to stressed levels throughout. The output is the number that matters: under this scenario, does coverage stay above 100% for every counterparty, and if not, what is the uncovered gap and the plan to close it? Regulators require this discipline explicitly — the CCP margin standards and bank collateralized-exposure frameworks both assume stressed, not point, valuations (see regulation).

Common Collateral Failures (and Their Antidotes)

Where these disciplines are applied at the largest scale — against every cleared derivative in the world — is the next page: collateral in derivatives.