Expected Credit Loss (ECL)

Expected credit loss (ECL) is the probability-weighted estimate of the cash a lender expects to lose on a loan, bond, receivable or other credit exposure over a given horizon. It is the measurement basis for loss allowances, and so for each impairment charge on credit exposures, under IFRS 9 Financial Instruments, which replaced the incurred-loss rules of IAS 39 for periods beginning on or after 1 January 2018. Banks, insurers, leasing companies and ordinary businesses holding accounts receivable all calculate ECL.

ECL is the number behind the allowance; the loan loss provision is the income-statement charge needed to move the allowance to that number. In practice ECL is usually built from three risk parameters, probability of default, loss given default and exposure at default. The US equivalent is the CECL model in ASC 326, which also uses expected losses but over the full life of every loan from day one.

Picture a landlord with a hundred tenants. Most pay on time, but every year a few fall behind, and some of those never pay what they owe even after the deposit is kept. The landlord can make a fair guess at how much rent will be lost next year by asking three questions: how many tenants are likely to stop paying, how much would each one owe at that point, and how much of it could be recovered?

Multiply those answers together and add them up, and the landlord has an estimate of expected losses. ECL is that same estimate done carefully for a loan book, with extra attention to whether the economy ahead looks better or worse than the past.

For a single exposure and a single period, the standard building-block formula is:

• $\text{PD}$ is the probability that the borrower defaults during the period.
• $\text{LGD}$ is the share of the exposure that would be lost if default happens, after recoveries.
• $\text{EAD}$ is the expected amount outstanding at the time of default, including likely drawdowns of undrawn limits.

A business loan of $1,000,000 with a 12-month PD of 2% and an LGD of 40% has a 12-month ECL of $0.02 \times 0.40 \times 1{,}000{,}000 = 8{,}000$, or $8,000. For a lifetime measure the lender repeats the calculation for each future year, using the marginal PD for that year and the expected balance, and discounts each result at the loan's effective interest rate before adding them up. IFRS 9 also requires the result to be weighted across several economic scenarios, such as a base, upside and downside case, rather than taken from a single forecast.

IFRS 9 changes the horizon of ECL according to how credit quality has moved since the loan was first recognised:

1. Stage 1. No significant increase in credit risk since origination. The allowance equals 12-month ECL, the lifetime losses that could result from defaults in the next 12 months. Interest is recognised on the gross carrying amount.
2. Stage 2. A significant increase in credit risk (SICR) has occurred, but the loan is not credit-impaired. The allowance jumps to lifetime ECL. The standard includes a rebuttable presumption that SICR has happened once payments are more than 30 days past due.
3. Stage 3. The loan is credit-impaired and is usually reported as a non-performing loan, which usually aligns with the lender's definition of default and a rebuttable 90-days-past-due presumption. Lifetime ECL applies and interest is recognised on the net carrying amount.

The move from Stage 1 to Stage 2 is often called the "cliff effect", because a modest worsening in a borrower's rating can multiply the allowance several times over. Trade receivables and lease receivables may use a simplified approach that always measures lifetime ECL, often through a provision matrix based on days past due.

ECL is not a mechanical output. Lenders must decide how to define SICR, which macroeconomic variables drive their PD models (for example unemployment, house prices or GDP), how many scenarios to use and how to weight them. When models cannot capture a new risk, such as the 2020 pandemic, banks add management overlays. Auditors and supervisors spend a great deal of time on these choices, and IFRS 7 requires disclosure of the stage balances, the movement in allowances and the key assumptions.

ECL for accounting differs from the expected loss used for regulatory capital under Basel III. The Basel version uses through-the-cycle PDs and downturn LGDs and a fixed one-year horizon, while IFRS 9 uses point-in-time, forward-looking estimates and lifetime horizons for Stage 2 and Stage 3 exposures.

Critics say the model can be procyclical, since allowances rise sharply as forecasts darken, reducing capital just as a downturn begins. The reliance on forecasts and overlays makes results hard to compare between banks, and the boundary between Stage 1 and Stage 2 varies by institution. Supporters reply that earlier recognition of losses gives investors and supervisors a more honest picture of credit risk than the old incurred-loss approach, which was blamed for delaying recognition of bad loans before the 2008 financial crisis. Firms also find that small data sets, especially for low-default portfolios such as lending to sovereigns or banks, make PD and LGD inputs unstable.

ECL estimates also feed other decisions, from loan pricing and underwriting standards to debt restructuring talks, where the lender compares the ECL of doing nothing with the loss from a negotiated deal.

• IFRS Foundation
• Basel Committee on Banking Supervision