PD, LGD, EAD Explained: ECL Inputs for Indian Banks
ECL = PD × LGD × EAD × Discount Factor. PD (Probability of Default) is the likelihood of default over the horizon. LGD (Loss Given Default) is the share not recovered after default — 1 minus the recovery rate. EAD (Exposure at Default) is the outstanding balance at the time of default. All three must be estimated using internal historical data, adjusted for macroeconomic scenarios, and computed on a Point-in-Time basis — not the Through-the-Cycle basis used in Basel capital models.
The Expected Credit Loss formula looks deceptively simple: ECL = PD × LGD × EAD × Discount Factor. But each of these three inputs requires its own modelling framework, historical data infrastructure, and forward-looking adjustment. Together they convert the abstract concept of “credit risk” into a rupee provision figure at the individual loan level. This article explains each component — what it means, how banks estimate it, and the specific requirements RBI has set for ECL computation.
The ECL Formula: How the Components Fit Together
For Stage 1, PD covers only the next 12 months. For Stage 2 and Stage 3, PD covers the full remaining life of the loan. The final ECL figure is then probability-weighted across multiple macroeconomic scenarios using Forward-Looking Information (FLI) before it becomes the provision number.
PD — Probability of Default
PD is the statistical likelihood that a borrower will default within a defined time period. “Default” under ECL is defined broadly — it includes 90 DPD but also other indicators of credit impairment such as fraud, insolvency proceedings, or restructuring that signals inability to pay.
Point-in-Time vs Through-the-Cycle PD
Smoothed across an entire economic cycle. Does not change much quarter-to-quarter. Used in Basel III regulatory capital calculations (IRB approach). Designed to be stable and conservative over long periods.
Reflects current economic conditions and the forward-looking macro outlook. Changes with each quarterly macro scenario update. A worsening GDP forecast increases PiT PD even if the borrower has not missed any payment. This is what RBI requires for ECL.
Banks that already use IRB models for Basel capital cannot simply reuse their TtC PD for ECL — they must recalibrate to PiT PD. This is a significant modelling exercise, particularly for banks that have never built PiT PD models before.
RBI Requirements for PD Modelling
- Minimum data history: 5–7 years of internal default data (longer for retail portfolios with low default rates)
- Segmentation: PD models must be built by meaningful segment — product type, borrower rating, geography, industry
- Macro linkage: PD must be linked to macroeconomic variables (GDP, sectoral indices, inflation) so it responds to scenario changes
- Horizon: 12-month PD for Stage 1; marginal period-PDs stitched together for lifetime PD in Stage 2/3
- Validation: PD models must be validated by an independent internal team; external validation every 3 years
PD in Practice: A Worked Example
A bank’s retail home loan PD model estimates the following for a salaried borrower with an LTV of 75%, a credit score of 720, and 3 years into a 20-year loan:
| Macro Scenario | 12-Month PiT PD | Scenario Weight | Weighted PD |
|---|---|---|---|
| Base (GDP 6.5%) | 0.40% | 60% | 0.24% |
| Optimistic (GDP 7.5%) | 0.22% | 20% | 0.04% |
| Pessimistic (GDP 4.5%) | 0.90% | 20% | 0.18% |
| Scenario-Weighted PD | 0.24% + 0.04% + 0.18% = 0.46% | ||
LGD — Loss Given Default
LGD measures how much of the outstanding exposure the bank will not recover if the borrower defaults. It is always expressed as a percentage of EAD:
If a bank lends ₹1 crore and, after a borrower defaults, recovers ₹65 lakh through property sale and legal action — the LGD is 35% (1 − 0.65). The ECL provision on that loan would include a 35% loss assumption on whatever EAD is outstanding.
What Drives LGD in Indian Banking
Residential property with LTV 60% has lower LGD than unsecured personal loans. Gold-backed loans have very low LGD due to liquid collateral.
India’s SARFAESI / DRT recovery process is slow — often 3–7 years. Longer recovery timelines increase LGD because future recoveries are discounted at EIR over a longer horizon.
Senior secured lenders recover more than subordinated or unsecured creditors. In consortium lending, pari passu sharing affects individual bank recovery rates.
Collateral values fall in downturns — property prices, machinery valuations, stock values. RBI requires Downturn LGD, estimated under stressed conditions, not average historical recovery rates.
Downturn LGD: The RBI Requirement
RBI requires LGD to be estimated on a Downturn basis — reflecting recovery rates during periods of economic stress rather than average conditions. This is the same concept used in Basel Advanced IRB approaches. The practical implication: if your average historical property recovery rate is 75%, but in the 2008–2009 or post-COVID stress period it fell to 55%, your Downturn LGD should be calibrated closer to 45% (1 − 55%), not the average 25%.
LGD by Loan Type: Indicative Ranges
| Loan Type | Typical Collateral | Indicative Downturn LGD |
|---|---|---|
| Home Loan (LTV 60–70%) | Residential property | 15–30% |
| Gold Loan | Gold jewellery / coins | 5–15% |
| Secured SME Term Loan | Commercial property / machinery | 30–50% |
| Unsecured Personal Loan / Credit Card | None | 60–85% |
| Large Corporate (Unsecured) | Pari passu / negative lien | 40–60% |
These are indicative ranges. Actual LGD must be estimated from the bank’s own historical recovery data, adjusted to Downturn conditions.
EAD — Exposure at Default
EAD is the total amount the bank is exposed to at the moment a borrower defaults. For most term loans, EAD is straightforward — it is the outstanding principal plus accrued interest at the expected default date. For revolving and contingent facilities, EAD must be modelled.
EAD for Term Loans vs Revolving Facilities
EAD = Outstanding principal + accrued interest at expected default date. For Stage 1, this is the projected balance 12 months from now. For Stage 2/3, it is the balance at each point in the future repayment schedule, probability-weighted and discounted.
EAD = Current drawn amount + (Undrawn limit × CCF). The Credit Conversion Factor (CCF) is the estimated fraction of the undrawn limit that will be drawn before default. Borrowers typically draw down their limits aggressively as they approach distress.
The CCF for revolving facilities is a particularly important modelling parameter in India, where working capital limits, cash credit accounts with drawing power, and overdraft facilities form a large part of bank lending to SMEs and corporates. Distressed borrowers tend to fully utilise their sanctioned limits before defaulting, making CCF close to 1.0 in many portfolios — meaning the full sanctioned limit should be treated as EAD, not just the current drawn amount.
The Discount Factor: Time Value in ECL
The discount factor converts future expected losses back to present value. RBI requires discounting at the loan’s Effective Interest Rate (EIR) — the rate that exactly equates the present value of all contractual cash flows to the initial carrying amount of the loan. For most plain-vanilla loans, EIR approximates the contracted interest rate. For loans with fees, embedded options, or modified cash flows, EIR must be computed separately.
The discount factor matters most for long-tenure loans. A ₹1 lakh loss expected 15 years from now, discounted at 9% EIR, has a present value of only ₹27,454. Ignoring this discount would overstate ECL provisions for long-dated assets.
Putting It All Together: A Full ECL Calculation
₹50 lakh
1.8%
45%
~0.92 (1 yr at 9% EIR)
= 0.018 × 0.45 × 50,00,000 × 0.92
= ₹37,260
Key Differences Between ECL Inputs and Basel IRB Inputs
| Parameter | Basel IRB (Capital) | ECL (Provisioning) |
|---|---|---|
| PD type | Through-the-Cycle (TtC) — stable, smoothed | Point-in-Time (PiT) — current conditions + macro FLI |
| LGD type | Downturn LGD (same concept) | Downturn LGD — adjusted for FLI scenarios |
| EAD horizon | 1 year (fixed) | 12 months (Stage 1) or full lifetime (Stage 2/3) |
| Macro adjustment | Implicit in TtC estimates | Explicit — probability-weighted scenario overlay required |
| Purpose | Regulatory capital (RWA calculation) | P&L provisioning (balance sheet) |
What is PD, LGD, and EAD in ECL?
PD (Probability of Default) is the statistical likelihood that a borrower will default within a defined time horizon. LGD (Loss Given Default) is the fraction of the exposure that the bank will not recover if default occurs — equal to 1 minus the recovery rate. EAD (Exposure at Default) is the total outstanding balance the bank is exposed to at the time of default. Together, ECL = PD × LGD × EAD × Discount Factor, computed at the individual loan level and adjusted for macroeconomic scenarios.
What is the difference between Point-in-Time PD and Through-the-Cycle PD?
Through-the-Cycle (TtC) PD is smoothed across an entire economic cycle and does not change significantly quarter-to-quarter — it is used in Basel IRB capital models. Point-in-Time (PiT) PD reflects current economic conditions and the forward-looking macro outlook, changing with each scenario update. RBI requires PiT PD for ECL — a worsening GDP forecast will increase PiT PD and therefore increase ECL provisions, even if no borrower has missed a payment.
What is Downturn LGD and why does RBI require it?
Downturn LGD is the loss given default estimated under stressed economic conditions — when collateral values are depressed and recovery timelines are longer than average. RBI requires Downturn LGD (rather than average historical LGD) to ensure ECL provisions are conservative and reflect realistic loss scenarios during economic downturns. For example, if average property recovery is 75% but falls to 55% in a downturn, Downturn LGD would be 45% (1 minus 55%), not the average 25%.
How is EAD calculated for revolving credit facilities?
For revolving facilities such as cash credit accounts, overdrafts, and credit cards, EAD = Current drawn amount + (Undrawn limit × Credit Conversion Factor). The CCF is the estimated fraction of the undrawn limit that a distressed borrower will draw before defaulting. Since borrowers in financial stress tend to maximise their credit lines before default, CCF is often close to 1.0 for revolving facilities — meaning the full sanctioned limit should effectively be treated as EAD.
Can banks reuse their Basel IRB models for ECL?
Not directly. Basel IRB uses Through-the-Cycle PD and fixed 1-year EAD horizons, whereas ECL requires Point-in-Time PD and lifetime EAD for Stage 2/3. Banks with IRB models must recalibrate their PD to PiT, extend their EAD modelling to cover the full loan life, and overlay probability-weighted macroeconomic scenarios. The Downturn LGD concept is shared between Basel and ECL, but the macro scenario overlay must still be applied for ECL purposes. Foreign banks with global IFRS 9 models will also need India-specific calibration.