Credit Risk Measurement: PD, LGD, EAD and VaR for CAIIB

Credit risk is the single largest risk a bank carries on its balance sheet. And measuring it accurately is the heart of the Risk Management elective for CAIIB. Mastering credit risk means understanding how lenders estimate the chance that a borrower defaults.

How much they would lose if that happens. And how those numbers roll up into capital requirements under the Basel framework. This guide walks CAIIB aspirants through the core measurement models — PD. LGD and EAD. Expected loss, and Value at Risk for market risk — in exam-ready language.

The Building Blocks: PD, LGD and EAD

Modern credit risk measurement rests on three parameters that every CAIIB candidate must know cold. Probability of Default (PD) is the likelihood that a borrower fails to meet obligations over a defined horizon, usually one year, expressed as a percentage. It is estimated from historical default data, internal rating grades or statistical scoring models. Loss Given Default (LGD) measures the proportion of exposure a bank actually loses after recovery efforts, net of collateral and guarantees; it is one minus the recovery rate. A fully secured loan may carry an LGD of 20 percent, while an unsecured personal loan can exceed 60 percent. Exposure at Default (EAD) is the amount outstanding when default occurs, which for term loans is close to the drawn balance but for credit lines includes a credit conversion factor on the undrawn portion. These three drivers feed directly into the regulatory capital formulas. A candidate who can define each parameter, state its units, and explain what raises or lowers it will handle a large share of the elective's numerical and conceptual questions. Remember the intuition: PD answers how likely, LGD answers how severe, and EAD answers how much. Better collateral pulls LGD down, a longer time horizon pushes PD up, and an actively used overdraft facility lifts EAD because more of the limit tends to be drawn as a borrower's finances deteriorate. Practising rated borrower scenarios on our CAIIB mock tests cements the difference between these measures.

Expected Loss and Unexpected Loss

Once PD, LGD and EAD are estimated, they combine into the most important formula in credit risk: Expected Loss (EL) = PD x LGD x EAD. Expected loss is the average loss a bank anticipates over a year and is treated as a cost of doing business, covered by provisions and priced into the interest rate charged to the borrower. For example, a loan with a one percent PD, 40 percent LGD and rupees 10 crore EAD carries an expected loss of rupees 4 lakh. But averages are not enough. Actual losses fluctuate around the expected value, and the deviation is called Unexpected Loss (UL) — the volatility of losses that provisions alone cannot absorb. Unexpected loss is what economic and regulatory capital must cover, because it represents the shocks that could threaten solvency. The distinction is central to the Basel philosophy: provisions buffer expected loss, while capital buffers unexpected loss. CAIIB questions frequently ask you to compute EL from given parameters or to explain why capital, not provisioning, is the correct cushion for unexpected loss. Keeping these two concepts cleanly separated is a high-yield exam skill that also reinforces the wider risk-management mindset. A useful mental picture is a loss distribution: expected loss sits at the peak as the most likely outcome, while unexpected loss is the distance from that peak out to a stressed percentile in the tail. Diversification across borrowers, sectors and geographies shrinks the tail and therefore the capital needed, which is why concentrated loan books attract supervisory attention. You can revise related capital concepts on our banking exam blog.

Value at Risk for Market Risk

While PD, LGD and EAD govern the loan book, banks also hold tradeable positions whose value swings with markets, and here the workhorse measure is Value at Risk (VaR). VaR estimates the maximum loss a portfolio is likely to suffer over a given holding period at a chosen confidence level. A one-day 99 percent VaR of rupees 5 crore means that on 99 of 100 trading days losses should not exceed rupees 5 crore; on roughly one day they might. The three classic computation methods are the variance-covariance (parametric) approach, which assumes normally distributed returns, the historical simulation method, which reprices the portfolio using past market moves, and the Monte Carlo simulation method, which generates thousands of random scenarios. Each balances accuracy against computational effort, and CAIIB candidates should be able to list the assumptions and limitations of each — VaR says nothing about the size of losses beyond the cutoff, which is why stressed VaR and expected shortfall supplement it. Although VaR addresses market risk rather than the loan book, the underlying idea of capturing a loss distribution at a confidence level is shared across both disciplines and recurs throughout the elective. Two parameters define any VaR figure and examiners love to test them: the holding period, which reflects how long it would take to unwind the position, and the confidence level, with regulators typically prescribing 99 percent for capital purposes. A longer holding period and a higher confidence level both enlarge the reported VaR. Banks also run back-testing, comparing actual daily losses against the VaR estimate to confirm the model breaches occur no more often than the confidence level implies. Reinforce these definitions with our risk terms matching game.

Internal Ratings and Basel Approaches

The Basel framework gives banks two broad routes to quantify credit risk for capital. Under the Standardised Approach, risk weights are prescribed by the regulator, largely driven by external credit ratings from approved agencies, and the bank simply maps each exposure to its assigned weight. Under the Internal Ratings-Based (IRB) Approach, supervised banks use their own estimates of the risk parameters. The IRB framework comes in two flavours: in the Foundation IRB, the bank estimates PD while the regulator supplies LGD and EAD, whereas in the Advanced IRB, the bank estimates PD, LGD and EAD itself, subject to strict validation. Internal rating systems assign each borrower to a grade that corresponds to a PD band, allowing differentiated pricing and capital, and a well-designed scale separates investment-grade names from sub-investment-grade ones with clear default rates for each rung. Indian banks have largely operated on the standardised approach, with the Reserve Bank of India governing the migration path and validation standards, including back-testing of internal estimates against realised defaults before an IRB model can be approved. CAIIB aspirants should know the boundary between what the bank estimates and what the regulator dictates under each approach. Track the latest prudential norms via our IIBF news updates and the official capital guidelines on the Reserve Bank of India website or the global standard-setter at the Bank for International Settlements.

Frequently Asked Questions

Conclusion

Credit risk measurement is best learnt as a connected chain: PD, LGD and EAD feed expected loss, the volatility around that average becomes unexpected loss requiring capital, VaR captures the same loss-distribution logic for market positions, and the Basel approaches decide who estimates what. Anchor every formula in the elective to this map and the numerical questions become routine. Ready to convert understanding into marks? Enrol in the CAIIB Risk Management course and attempt a full-length CAIIB mock test today to lock in these high-yield concepts.