CAIIB credit risk models: PD, LGD and EAD

For the CAIIB Risk Management elective, credit risk models are the analytical backbone of how a bank measures the chance a borrower defaults and how much money it stands to lose. Whether you are appraising a working-capital limit for an MSME or building a portfolio view for the board's Risk Management Committee, the same three quantities keep recurring: Probability of Default (PD), Loss Given Default (LGD) and Exposure at Default (EAD). Get comfortable with these and the rest of the Basel capital story falls into place. This guide walks through each component, the expected-loss equation that binds them, the internal rating framework banks use in India, and how stress testing pressure-tests the whole system.

What Credit Risk Models Measure: PD, LGD and EAD

Credit risk models exist to convert a fuzzy question - "how risky is this loan?" - into numbers a bank can price, provision and hold capital against. Three parameters do the heavy lifting. Probability of Default (PD) is the likelihood that a borrower fails to meet its obligations over a one-year horizon; in India a default is broadly aligned with the 90-days-past-due Non-Performing Asset (NPA) definition used by the Reserve Bank of India. Loss Given Default (LGD) is the share of exposure a bank actually loses after recoveries, collateral realisation and legal proceedings - so an LGD of 45% means the bank recovers 55 paise on the rupee. Exposure at Default (EAD) estimates how much will be outstanding at the moment of default, which matters hugely for revolving facilities like cash credit where a stressed borrower tends to draw down the full sanctioned limit.

These are not academic abstractions. A retail home-loan book will show low PD and low LGD because of tangible mortgage security, while an unsecured personal-loan portfolio carries high PD and high LGD. Understanding the interplay is central to the CAIIB syllabus and to real risk practice. If you want to test how well the definitions have stuck, try a quick round on the match game before moving on.

Expected Loss and Unexpected Loss

Once you have the three parameters, the headline formula of credit risk modelling is elegantly simple: Expected Loss (EL) = PD x LGD x EAD. If a bank estimates a 2% PD, a 40% LGD and an EAD of Rs 1 crore, the expected loss is 0.02 x 0.40 x 1,00,00,000 = Rs 80,000. Expected loss is treated as a cost of doing business - it is covered through provisioning and priced into the loan's interest spread, not through regulatory capital.

Capital, by contrast, exists to absorb Unexpected Loss (UL) - the volatility of losses around that expected average in a bad year. A portfolio can breeze through several benign years and then suffer a spike when a sector like real estate or an infrastructure cluster turns down together. Because defaults are correlated, UL is far larger than a naive sum of individual variances would suggest, which is why concentration risk is watched so closely. The Basel framework calibrates minimum capital to a 99.9% confidence level over one year, meaning a bank should survive all but a one-in-a-thousand-year loss event.

• Expected Loss - covered by provisions and loan pricing.
• Unexpected Loss - covered by regulatory and economic capital.
• Stress/tail loss - explored through scenario and stress testing, discussed below.

Grasping the EL versus UL split is one of the most-examined distinctions for CAIIB candidates. Reinforce it with the practice sets on our test series, which mirror the calculation style of the actual paper.

Basel III, IRB and Rating Models

The regulatory home of these parameters is the Basel framework. Under Basel norms banks may compute credit-risk capital using the Standardised Approach, where risk weights are driven by external ratings from agencies such as CRISIL, ICRA, CARE and India Ratings, or the Internal Ratings-Based (IRB) Approach, where the bank estimates PD (Foundation IRB) or PD, LGD and EAD together (Advanced IRB) using its own models. Basel III strengthened this architecture after the 2008 crisis by raising the quality and quantity of capital, adding the Capital Conservation Buffer, the Countercyclical Buffer and a leverage ratio, and layering on liquidity standards (LCR and NSFR). RBI implements these standards in India with its own timelines and add-ons.

At the coalface, rating models assign each borrower to a grade on an internal rating scale, and each grade maps to a PD band. Corporate models blend financial ratios (leverage, interest cover, current ratio), qualitative factors (management quality, industry outlook) and behavioural signals (past repayment conduct). Retail portfolios lean on statistical scorecards - logistic regression or, increasingly, machine-learning models scored on application and bureau data from CIBIL and other credit information companies. Whatever the technique, RBI and Basel demand rigorous model validation: discrimination tests, calibration checks and back-testing of predicted versus realised default rates. Keep an eye on the latest circulars via IIBF news and current policy rates on our RBI rates tracker.

Stress Testing the Credit Portfolio

Even the best-calibrated credit risk models describe the world as it usually behaves. Stress testing asks the harder question: what happens under severe but plausible shocks? Banks run sensitivity analysis (flexing one variable, say a 200-basis-point rate rise) and scenario analysis (a coherent macro downturn - GDP contraction, rupee depreciation, a commodity-price collapse) and observe how PDs migrate upward, LGDs worsen as collateral values fall, and capital ratios come under pressure. RBI mandates stress testing as part of the supervisory framework and publishes system-wide results in its half-yearly Financial Stability Report.

A closely related exercise is reverse stress testing, which starts from the point of failure and works backwards to identify the scenarios that would break the bank - useful for spotting hidden concentrations. The output feeds the Internal Capital Adequacy Assessment Process (ICAAP), where a bank justifies holding capital above the regulatory minimum. For an exam candidate, the key takeaway is that stress testing links the micro parameters (PD, LGD, EAD) to the macro resilience of the institution, closing the loop between measurement and management. Browse more worked explanations on our blog and, if you are also sitting the foundation exams, revise the fundamentals through the JAIIB track.

Frequently Asked Questions

Conclusion

Credit risk models turn the vague fear of default into disciplined numbers: PD, LGD and EAD combine into expected loss, capital cushions the unexpected loss, Basel III and IRB set the rules, rating models grade the borrowers, and stress testing checks the whole structure against a storm. Master these connections and the CAIIB Risk Management elective becomes far more intuitive. Ready to put it to the test? Attempt a full-length mock on our CAIIB test series or deep-dive the syllabus through the CAIIB course and lock in your score.