Credit Risk Models Explained: PD, LGD and EAD for IIBF

For any banker preparing for the IIBF Risk Management certification. Mastering credit risk models is non-negotiable. These quantitative frameworks estimate how likely a borrower is to default. How much a lender stands to lose. And how much exposure is at risk when default strikes.

Modern credit risk models rest on three pillars: Probability of Default (PD). Loss Given Default (LGD) and Exposure at Default (EAD). This guide breaks down each component.

Links them to Basel III capital rules. And shows how rating models. Scorecards and credit Value-at-Risk fit together in a working risk framework.

What Credit Risk Models Actually Measure

Credit risk is the danger that a borrower fails to honour contractual obligations. Banks cannot simply hope this away; they quantify it. The core identity every candidate must memorise is the Expected Loss equation. Where the three parameters combine multiplicatively.

• PD (Probability of Default). The likelihood a borrower defaults within a one-year horizon. Expressed as a percentage.
• LGD (Loss Given Default) — the fraction of exposure not recovered after default. Net of collateral and recoveries.
• EAD (Exposure at Default). The outstanding amount a lender is exposed to at the moment of default. Including drawn balances and a portion of undrawn limits.

Putting these together gives the headline formula: Expected Loss = PD x LGD x EAD. Unexpected loss, the volatility around this average, is what regulatory capital is designed to absorb. Robust models separate expected loss (a cost of doing business, covered by provisions) from unexpected loss (covered by capital). Understanding this split is the foundation on which everything else in the Basel framework is built, and it appears repeatedly in IIBF examinations. Practising application questions on our mock tests reinforces how each parameter feeds the final capital number.

PD, LGD and EAD in Detail

Each parameter is estimated differently, and examiners love probing the distinctions. PD is typically derived from internal rating grades or statistical scorecards calibrated to historical default frequencies. A borrower rated AAA-equivalent might carry a PD far below one percent.

While a sub-investment-grade obligor carries materially higher odds. LGD depends heavily on seniority. Collateral: a fully secured home loan recovers more than an unsecured personal loan.

So its LGD is lower. EAD captures the credit-conversion factor applied to off-balance-sheet commitments such as undrawn overdrafts. Guarantees.

Good credit risk models keep these three parameters statistically independent where possible. Then recombine them. Banks back-test each parameter against realised outcomes during validation. Ensuring PD estimates match observed default rates. LGD assumptions match actual recovery experience.

Standardised vs IRB Approaches Under Basel

Basel III offers two broad routes for computing credit-risk capital. And the choice shapes how much modelling a bank does in-house. Under the Standardised Approach.

Risk weights are prescribed by the regulator. Often tied to external credit ratings. So the bank does minimal internal estimation.

Under the Internal Ratings-Based (IRB) approach. The bank uses its own credit risk models to estimate the risk parameters. Subject to supervisory approval.

• Standardised Approach — regulator-set risk weights; simplest; no internal PD/LGD estimation required.
• Foundation IRB (F-IRB) — the bank estimates PD internally. While LGD and EAD use supervisory values.
• Advanced IRB (A-IRB) — the bank estimates PD. LGD and EAD with its own validated models.

The IRB route can be more capital-efficient for well-managed portfolios, but it demands rigorous governance, multi-year data histories and independent validation. The Reserve Bank of India and the Basel Committee set the qualifying criteria; candidates should review the framework on the RBI website and track regulatory updates through our IIBF news feed. Migrating from standardised to advanced status is a multi-year supervisory journey, not a switch a bank flips overnight.

Rating Models, Scorecards and RAROC

The engine room of credit risk management is the rating model. Corporate exposures are usually assessed with expert-judgement-plus-financials rating models that map borrowers onto an internal grade scale. Retail exposures.

By contrast. Are scored through statistical scorecards built on application and behavioural data. Logistic-regression credit risk models that output a score convertible into a PD.

Both feed the same Expected Loss machinery.

Banks then layer pricing discipline on top using RAROC. Risk-Adjusted Return on Capital. RAROC divides risk-adjusted return by the economic capital a deal consumes.

Ensuring that loans are priced to compensate for their expected loss. Capital cost. A facility that looks profitable on a gross-margin basis may destroy value once its PD.

LGD and capital charge are factored in.

• Rating models — grade corporate and SME borrowers using financials and qualitative factors.
• Scorecards — statistically rank retail applicants and existing accounts.
• RAROC — aligns loan pricing with the risk-adjusted capital each exposure absorbs.

Sharpen your recall of these terms with our quick match game before exam day.

Credit VaR, CreditMetrics and Portfolio Risk

Single-name expected loss is only half the story. At the portfolio level. Banks deploy credit Value-at-Risk (Credit VaR) to estimate the maximum loss not exceeded at a given confidence level over a chosen horizon.

Frameworks such as CreditMetrics model not just defaults but also rating migrations. The chance that exposures shift to better or worse grades. And use correlations to capture how borrowers default together in a downturn.

Concentrated or highly correlated portfolios carry far more unexpected loss than diversified ones.

This is where through-the-cycle (TTC) versus point-in-time (PIT) calibration matters. TTC ratings average risk across an economic cycle and are more stable. Suiting capital planning.

PIT estimates reflect current conditions and react quickly to stress. Suiting provisioning and early-warning systems. Strong credit risk models make this choice explicitly and document it.

• Credit VaR — portfolio loss at a confidence level, capturing default correlation.
• CreditMetrics — migration-and-correlation framework for portfolio credit risk.
• TTC vs PIT — stable cyclical view versus responsive current-conditions view.

Keep an eye on policy rates that influence default conditions via our RBI rates tracker.

Model Validation and Governance

A model is only as trustworthy as its validation. Supervisors require that credit risk models be independently validated. Regularly back-tested and benchmarked against alternatives.

Validation checks discrimination (does the model rank good and bad borrowers correctly?). Calibration (do predicted PDs match realised default rates?). Stability (do scores hold up over time?).

Documentation, data quality controls and a clear model-risk-management policy are equally essential.

Governance failures, not mathematics, cause most model breakdowns. Stale data. Unmonitored overrides.

Untested assumptions quietly erode accuracy until a downturn exposes them. Boards must ensure an independent validation unit. Periodic review cycles and conservative add-ons where data is thin.

The Basel framework. RBI guidelines both insist on this discipline before granting IRB permission.

For more study walkthroughs, browse our risk-management blog.

Frequently Asked Questions

Conclusion

Credit risk models translate uncertainty into numbers a bank can price, provision and capitalise against. From the Expected Loss identity through PD, LGD and EAD, to standardised versus IRB approaches, rating models, scorecards, RAROC, Credit VaR and rigorous validation, the IIBF Risk Management syllabus rewards candidates who connect these pieces into one coherent framework. Ready to test your command of these concepts under exam conditions? Put your knowledge to work on our practice papers at iibf.store/tests and walk into the exam hall with confidence.