Expected credit loss: probability, exposure, and recovery

Expected credit loss is the probability-weighted amount a lender expects to lose when borrowers fail to perform. A simplified estimate multiplies probability of default, loss given default, and exposure at default. It is an average loss estimate over specified conditions and time, not the largest loss the lender can suffer.

Probability, severity, and exposure answer different questions

Probability of default, or PD, estimates the chance that the borrower meets the defined default condition within a stated horizon. A one-year PD cannot be substituted for a lifetime PD.

Loss given default, or LGD, estimates the fraction of exposure lost conditional on default. Recoveries, collection costs, collateral enforceability, and the time required to collect affect this measure. A 40% LGD means a 40% loss on the specified exposure after the assumed recovery treatment.

Exposure at default, or EAD, estimates the amount outstanding when default occurs. It can differ from today’s drawn balance when a borrower has an undrawn commitment or when contractual balances change before default.

Expected loss combines the three inputs

Take EAD of $1 million, a one-year PD of 2%, and LGD of 40%. The simplified expected loss is $1,000,000 × 0.02 × 0.40 = $8,000 over the stated year.

That result does not predict an $8,000 loss on this borrower. Under a simplified two-outcome model, the borrower either avoids default or defaults with a $400,000 loss. The $8,000 is the probability-weighted average of those outcomes.

For several exposures, expected losses can be added when the inputs refer to compatible horizons and definitions. Adding the averages does not describe the distribution of the portfolio’s total loss.

Concentration changes the severe-loss outcome

Two portfolios can have the same expected loss and different exposure to a common shock. Loans concentrated in one employer, property market, or industry can default together.

Dependence between defaults affects the frequency and size of aggregate losses even when individual expected losses are unchanged. In a recession, both default probabilities and recovery assumptions can also deteriorate, changing the expected losses themselves.

Capital, stress testing, and concentration controls therefore address questions that the average-loss calculation does not answer.

Accounting allowances require their own framework

The PD × LGD × EAD expression is a compact risk model. An accounting allowance can require multiple scenarios, discounting, changing exposures, and a horizon determined by the applicable standard.

The reporting record must identify the default definition, horizon, exposure basis, recovery costs, and valuation date. Calling two results “expected loss” does not make them comparable when those inputs differ.