Credit Risk Models PD LGD EAD: A Complete Guide for Bankers

Every credit decision a bank makes rests on three quiet numbers working behind the scenes. When a CAIIB candidate first meets credit risk models PD LGD EAD, the acronyms can feel abstract, but they describe something every branch manager already senses instinctively: how likely is this borrower to default, how much will we lose if they do, and how much money is actually on the line at that moment? Regulators formalised this intuition into a rigorous framework because ad-hoc credit judgement does not scale across a portfolio of lakhs of accounts, and supervisors need a consistent way to size the capital a bank must hold against its loan book. This article walks through the three parameters, shows how they combine into expected loss, compares the regulatory approaches banks can use to estimate them, and flags the governance issues examiners love to test.

📊 Breaking Down PD, LGD and EAD

Probability of Default (PD) is the likelihood, over a one-year horizon, that a borrower will fail to meet contractual obligations. Banks estimate PD using historical default rates grouped by rating grade, scorecards built on financial ratios, or statistical models such as logistic regression trained on years of loan performance data. A retail home-loan borrower with a stable salary and low loan-to-value ratio sits in a low-PD bucket; an unsecured personal loan to a thin-file borrower sits much higher.

Loss Given Default (LGD) answers a different question: once default has actually happened, what fraction of the exposure does the bank fail to recover after accounting for collateral realisation, guarantees and recovery costs? LGD is expressed as a percentage of exposure and is the mirror image of the recovery rate — an LGD of 40% implies the bank expects to eventually recover 60 paise on every rupee outstanding. Collateral quality drives this number more than any other factor; the chapter on regulatory capital and capital adequacy ties LGD directly into how much regulatory capital a loan consumes.

Exposure at Default (EAD) is simply the amount the bank stands to lose at the moment of default — straightforward for a fully drawn term loan, but trickier for a cash-credit limit or an undrawn overdraft, where borrowers tend to draw down further just before trouble hits. EAD therefore adds an estimate of additional drawdown to the currently outstanding balance.

💡 Exam Tip: Remember the sequence PD → LGD → EAD → Expected Loss. Examiners frequently swap definitions between LGD and EAD in distractor options, so anchor each term to its one defining question — likelihood, loss fraction, or amount at stake.

🧮 From Risk Parameters to Expected Loss

Once a bank has estimated all three parameters for an exposure, it combines them using a formula every risk management student must be able to reproduce instantly: Expected Loss (EL) = PD × LGD × EAD. This single number represents the average loss a bank should anticipate on that exposure over the coming year, and pooled across a portfolio it becomes the basis for loan-loss provisioning under expected-credit-loss accounting frameworks.

Consider a working example. A bank holds an exposure at default of Rs 100 crore against a borrower with a 2% probability of default and an expected recovery rate of 60% on the collateral pool, giving an LGD of 40%. Expected loss works out to 0.02 × 0.40 × Rs 100 crore = Rs 80 lakh. That figure feeds two separate downstream calculations: the provisioning charge taken against the profit and loss account, and the risk weight used to compute regulatory capital. Because provisions and capital both trace back to the same three inputs, a bank that underestimates PD or LGD ends up both under-provisioned and under-capitalised at the same time — a combination examiners like to probe.

Unexpected loss, the volatility around this average figure, is what regulatory and economic capital are actually meant to absorb; expected loss is meant to be covered by pricing and provisions, not capital. This distinction between the two loss concepts is one of the more commonly confused ideas in the syllabus, and getting it right is what separates a strong RM score from an average one.

🏦 Standardised vs IRB: Who Estimates What

Basel's credit risk capital framework gives banks a choice of three approaches, each demanding progressively more sophistication and, in return, allowing progressively more risk sensitivity. Under the Standardised Approach, PD, LGD and EAD are effectively not modelled by the bank at all — risk weights are prescribed using external credit ratings and supervisory tables. Under Foundation IRB (F-IRB), a bank builds and validates its own PD models but still uses supervisor-prescribed LGD and credit conversion factors for EAD. Under Advanced IRB (A-IRB), a sufficiently mature bank estimates all three parameters — PD, LGD and EAD — from its own internal data and models, subject to strict supervisory approval and ongoing validation.

India's larger scheduled commercial banks operate mostly under the Standardised Approach today, with RBI permitting IRB adoption only for banks that meet strict data-history, validation and governance prerequisites laid out in RBI's guidelines on the Internal Ratings-Based approach. Moving up this ladder is attractive because internally modelled parameters are usually more risk-sensitive and can lower capital charges for genuinely low-risk books, but the trade-off is a heavier compliance and data burden. The same underlying logic — matching capital to measured risk rather than a blunt average — also underpins market-risk regimes like the Fundamental Review of the Trading Book, which candidates should study as a parallel track to credit risk modelling.

🔍 Data, Governance and Model Risk

A PD/LGD/EAD model is only as good as the data and oversight behind it. Banks need long historical default and recovery time series segmented by product and geography, and models must be back-tested regularly against actual outcomes rather than assumed to remain accurate forever. Weak governance around who signs off on model changes, how overrides are documented, and how frequently parameters are recalibrated is one of the most common findings in supervisory inspections, and it links directly to the wider theme covered in the corporate governance portion of the syllabus, where board-level ownership of risk models is emphasised.

Collateral valuation deserves special mention because it feeds LGD directly. A stale valuation, an over-optimistic haircut, or a legal defect in the security documentation can silently understate LGD for years until a default finally exposes the gap. This is exactly the terrain covered in the sibling article on collateral management and haircuts, which examines how conservative haircut policies keep LGD estimates honest.

Finally, these parameters do not exist in isolation — they feed pricing decisions too. A bank comparing two credit proposals with similar headline yields but different PD/LGD/EAD profiles should evaluate them using risk adjusted return on capital rather than gross spread, since RAROC explicitly weights return by the expected loss and capital consumed. Regulators require this discipline precisely because banking exists to intermediate risk profitably rather than blindly, a point also made in the foundational chapter on why do banks need regulation. Credit exposures also interact with capital markets activity — for instance, securities held as collateral often trade through the depositories covered in the JAIIB elective on stock exchanges and depositories in India, which is worth a cross-reference for candidates studying both papers.

⚠️ Common Mistake: Students often assume LGD is a fixed constant for a loan type. In reality LGD moves with the economic cycle — recoveries fall and haircuts widen during downturns, which is why "downturn LGD" estimates are mandated for regulatory capital purposes.

📌 Remember: EL = PD × LGD × EAD is portfolio-level provisioning math; unexpected loss above this average is what capital, not provisions, is meant to cover.

🧠 Practice MCQs: Credit Risk Models PD LGD EAD

Want chapter-wise mock tests with 100+ MCQs? Start practising free →

❓ Frequently Asked Questions

Mastering PD, LGD and EAD is less about memorising a formula and more about understanding how a handful of estimated parameters ripple through provisioning, pricing and regulatory capital across an entire loan book. For IIBF Risk Management candidates, this topic rewards conceptual clarity over rote learning — practise translating a case-study scenario into the EL formula until it becomes automatic. If you are preparing for the CAIIB elective, reinforce this chapter with structured mock tests on iibf.store's CAIIB course and revisit the full Risk Management topic hub for related chapters on capital, governance and operational risk.