Learn / Regulation

Expected credit loss explained: PD, LGD, EAD, Ind AS 109 and the RBI framework

By the Bucketwise team · 9 min read · Updated 13 November 2026

Expected credit loss, or ECL, is an estimate of what a lender expects to lose on a loan, replacing a fixed percentage by asset category. It is built from three numbers per loan, probability of default, loss given default and exposure at default, and one rule about which stage the loan is in.

How is expected credit loss calculated?

ECL = PD x LGD x EAD.
PD, probability of default: the chance the borrower defaults in the period.
LGD, loss given default: the share of the exposure lost after recoveries and collateral.
EAD, exposure at default: the amount outstanding if the default happens.

Illustrative loan, figures in crore
PD over the next 12 months	2%
LGD	45%
EAD	100
ECL	0.02 x 0.45 x 100 = 0.9

Real models add discounting and forward-looking adjustments. The multiplication is the core, and it is the part an analyst builds from roll rates, vintages and recovery data.

What are the three ECL stages?

Stage	Test	Loss measured
Stage 1	No significant increase in credit risk since initial recognition	12-month ECL
Stage 2	Credit risk has increased significantly, but the asset is not credit-impaired	Lifetime ECL
Stage 3	Credit-impaired at the reporting date	Lifetime ECL

RBI's bank directions set out the three stages in the same way (para 25 of the Directions of 27 April 2026). Stage 3 is where the NPA book sits, under the 90-day test in the classification norms. The directions state the 90-day test for NPAs and do not say in so many words that Stage 3 and NPA are the same set, so check how your lender maps them.

How does a loan move between stages in Ind AS 109?

Ind AS 109 follows IFRS 9. IFRS 9 presumes, unless the lender can show otherwise, that credit risk has increased significantly once contractual payments are more than 30 days past due. It also uses a rebuttable presumption that default occurs no later than 90 days past due. A lender can use a different trigger only with reasonable and supportable information.

This is why DPD and buckets are still the base of most staging rules. A loan in the 31-60 bucket is past the 30-day presumption, so it is a Stage 2 candidate unless the lender can rebut the presumption.

Worked example: ECL across the three stages

Three illustrative loans of 100 crore each, LGD 40% for Stage 1 and 2
Stage 1: 12-month PD 1.5%	0.015 x 0.40 x 100 = 0.6 crore
Stage 2: lifetime PD 10%	0.10 x 0.40 x 100 = 4.0 crore
Stage 3: PD 100%, LGD 55%	1.00 x 0.55 x 100 = 55.0 crore

The Stage 2 loan carries almost seven times the Stage 1 provision, though nothing has been missed yet. That jump is the point of the staging rule. It makes lenders recognise the loss when risk rises, not when payments stop.

All figures are made up. Real PDs and LGDs come from the lender's own data and models.

What does RBI change for banks from 1 April 2027?

The Directions are the Reserve Bank of India (Commercial Banks, Asset Classification, Provisioning and Income Recognition) Directions, 2026, issued on 27 April 2026. They take effect on 1 April 2027.
They apply to commercial banks. Small finance banks, payments banks and local area banks are outside the definition, and NBFCs are not covered.
The existing directions on income recognition, asset classification and provisioning are repealed when the new ones start.
Until then the 28 November 2025 directions apply. The provisioning norms guide has those rates.

The new directions set minimum provisions for Stage 1 and Stage 2 by category. Examples from para 82 are below.

Category	Stage 1 minimum	Stage 2 minimum
Secured retail loans	0.40%	5%
Corporate loans	0.40%	5%
Loans to micro and small enterprises	0.25%	5%
Farm credit	0.25%	5%
Housing loans to individuals	0.25%	1.50%
Gold loans	0.40%	1.50%
Unsecured retail loans	1%	5%

Stage 3 minimums step up with the time the asset has spent in Stage 3. In the categories we read, they reach 100% after four years. The table has more categories and more rows than this one. Read para 82 for yours.

The transition

The extra provision on 1 April 2027 does not go through the profit and loss account. It is charged to opening earnings, and banks may add the amount back to CET1 capital, net of tax, in declining fractions of four-fifths, three-fifths, two-fifths and one-fifth over the four years ending 31 March 2031 (paras 107 and 108). The first ECL-based reporting covers 30 June 2027 (para 147).

How does this apply to NBFCs?

NBFCs that follow Ind AS already hold ECL under Ind AS 109. The RBI directions for NBFCs act as a floor, and a shortfall against the floor goes to an impairment reserve. The bank ECL directions above do not apply to them.

What does an analyst build for ECL?

PD by segment, from roll rates and vintage curves.
LGD from recoveries on defaulted loans, by product and by security type.
EAD from outstanding plus expected drawdown, for limits.
Staging flags built from DPD, restructuring and watch-list status.
A reconciliation from the model's allowance to the finance ledger.

Sources

Try it on your own tables

The free tools give you the prompt, the checks and how to present the result. You enter table names only.

See the free tools

Frequently asked questions

What is expected credit loss?

The loss a lender expects on a loan: the probability of default times the loss given default times the exposure at default, with a 12-month or lifetime horizon depending on the stage.

What are the three stages of ECL?

Stage 1 for loans with no significant increase in credit risk (12-month ECL), Stage 2 for loans with a significant increase (lifetime ECL), and Stage 3 for credit-impaired loans (lifetime ECL).

When does the RBI ECL framework start?

For commercial banks, the Directions issued on 27 April 2026 take effect on 1 April 2027. Small finance banks, payments banks, local area banks and NBFCs are outside them.

Does Ind AS 109 use 30 and 90 days past due?

It presumes a significant increase in credit risk at more than 30 days past due and default at 90 days past due. Both can be rebutted with supportable information.

Keep reading

Regulation