Expected Credit Loss for Interviews: Explain PD, LGD, EAD and Provisioning Clearly
A lender can look profitable on Monday and look under-provisioned on Friday - not because cash disappeared, but because the expected future loss became visible. That is the shift Expected Credit Loss created: credit risk moved from a backward-looking accounting entry to a forward-looking management decision.
- Expected Credit Loss (ECL) is the probability-weighted estimate of credit losses, recognized before actual default.
- The core formula is ECL = PD × LGD × EAD, adjusted for time value and forward-looking scenarios.
- PD asks: “How likely is default?” LGD asks: “If default happens, how much will we lose?” EAD asks: “How much exposure will be outstanding?”
- Under IFRS 9 / Ind AS 109, loans move across Stage 1, Stage 2 and Stage 3 as credit risk deteriorates.
- Provisioning is not just accounting - it affects profit, capital, lending appetite, pricing and investor confidence.
- The best interview answer combines math, accounting treatment, business judgment and risk governance.
- The trap: treating ECL as a simple historical NPA percentage instead of a forward-looking risk estimate.
Think of ECL as a lender’s credit-risk dashboard. It links a borrower’s chance of default, the likely loss if default happens, and the amount the lender is exposed to - then converts that risk into a provision in the financial statements.
Core Explanation: The Three Inputs Behind ECL
Expected Credit Loss is not “how much default happened last year.” It is a probability-weighted estimate of future credit loss on current financial assets - loans, bonds, trade receivables, lease receivables and some loan commitments.
The simplified formula is:
ECL = Probability of Default × Loss Given Default × Exposure at Default
In real banking models, this is refined using multiple macroeconomic scenarios, discounting to present value, behavioural assumptions, collateral values and portfolio segmentation.
Definitions You Must Be Able to Say in One Breath
IFRS 9 defines expected credit losses as “the weighted average of credit losses with the respective risks of a default occurring as the weights.”
Why ECL Replaced the Old Comfort of “Wait Until Loss Happens”
The older incurred-loss model recognized provisions mainly when there was objective evidence of impairment. The weakness became obvious during credit cycles: provisions often came too late, making reported profits look better before the stress surfaced.
ECL changes the timing. It asks lenders to recognize credit deterioration earlier, using reasonable and supportable forward-looking information.
The IFRS 9 / Ind AS 109 Staging Model
IFRS 9 and India’s Ind AS 109 use a three-stage impairment approach. The stage determines whether the lender records a 12-month expected loss or a lifetime expected loss.
Worked Example: Calculate ECL in 60 Seconds
Assume a bank has a ₹10,00,000 unsecured business loan outstanding.
- PD = 4%
- LGD = 60%
- EAD = ₹10,00,000
ECL = 4% × 60% × ₹10,00,000 = ₹24,000
If the borrower’s risk worsens and the account moves to Stage 2, the PD may be estimated over the loan’s lifetime rather than only the next 12 months. Suppose lifetime PD becomes 12% while LGD and EAD remain unchanged:
Lifetime ECL = 12% × 60% × ₹10,00,000 = ₹72,000
The business lesson is simple: the borrower has not defaulted yet, but the provision has increased by ₹48,000 because expected risk increased.
Metrics a Credit Analyst Tracks
Do not stop at the formula. In a bank, ECL quality is judged using model metrics, portfolio metrics and accounting ratios together.
Mini Case Study: Bajaj Finance and ECL as a Living Risk System
Bajaj Finance shows how a large Indian NBFC can use expected-loss thinking to price, approve, monitor and provide for consumer credit at scale.
Situation: Bajaj Finance operates in a market where consumer loans, personal loans, EMI finance and cross-sell products can grow quickly, but borrower risk can change just as quickly. In India, NBFCs reporting under Ind AS 109 must recognize expected credit losses, while RBI regulation and supervisory actions shape underwriting, provisioning discipline and capital allocation.
The move: The company’s risk approach is not just “lend more to good customers.” The primary driver is granular segmentation - using customer repayment history, bureau behaviour, product type and risk scorecards to decide approval, pricing, limits and collections intensity. Supporting drivers include diversified loan products, strong collection infrastructure, portfolio monitoring, conservative provisioning practices and response to RBI signals such as tighter treatment of unsecured consumer credit risk.
The lesson: ECL is most powerful when it is embedded before disbursement, not calculated only at quarter-end. A lender that understands PD, LGD and EAD by product and segment can slow risky growth, price credit better and recognize stress earlier.
So what: Bajaj Finance is a useful interview example because it proves ECL is not an accounting footnote. It is a management operating system linking customer selection, product design, pricing, collections and profitability.
How AI Changes Expected Credit Loss
AI is changing ECL in practical, model-level ways - especially for lenders with large retail, MSME or digital portfolios.
Student workflow: Use NotebookLM or ChatGPT with a bank or NBFC annual report. Ask: “Extract the ECL policy, staging rules, credit cost trend, Stage 2 movement and management commentary. Convert it into a 90-second interview answer.” Then verify every number against the annual report before using it.
AI can help identify risk patterns, but ECL models still require explainability, validation, bias checks, audit trails and compliance with accounting standards and RBI expectations.
Interview Relevance
“Explain Expected Credit Loss. How are PD, LGD and EAD used, and why did IFRS 9 make provisioning more forward-looking?”
If asked a numerical follow-up, write the formula first, calculate clearly, then explain what would change if the account moved from Stage 1 to Stage 2.
Common Mistake
The costly mistake is saying ECL is “a provision based on past NPAs.” That misses the entire point: ECL is forward-looking and probability-weighted. The fix: always say “PD × LGD × EAD, adjusted for staging and future scenarios.”
What to Revise Next
Now move from estimating expected loss to understanding what happens when credit actually goes bad. Revise Distressed Debt & Recovery: What Happens When Credit Goes Bad, then apply the full lending logic in Case Study: Would You Lend? A Full Credit Decision Walkthrough.