Why is loss given default important? Because LGD determines the actual financial loss a bank absorbs when a borrower defaults. It feeds directly into capital requirements under Basel III, provisioning under IFRS 9 and RBI’s ECL framework, and loan pricing decisions. As a result, getting LGD wrong means a bank either holds too little capital against real losses, or misprices credit risk across its portfolio.
LGD measures the share of an exposure a lender doesn’t recover after a borrower defaults. In other words, it’s the loss severity component of credit risk. That makes it distinct from Probability of Default (PD), which measures how likely a default is in the first place.
Together with Exposure at Default (EAD), these three components combine into the standard credit loss formula:
Expected Loss = PD × LGD × EAD
This formula is really the starting point for understanding why is loss given default important in practice. PD tells a bank how often defaults happen. LGD, on the other hand, tells the bank how much each one costs. So a portfolio with a low default rate but high LGD can generate the same expected loss as a portfolio with a high default rate but low LGD. That’s exactly why risk teams can’t focus on PD alone and call the job done. For a deeper walkthrough, see our companion guide on LGD recovery rates and estimation.
Basel III uses LGD as a direct input into risk-weighted assets (RWA). This, in turn, determines how much regulatory capital a bank must hold against a given exposure. Under the Foundation IRB approach, for instance, supervisors prescribe standard LGD values: 45% for senior unsecured corporate exposures, and 75% for subordinated claims. Under the Advanced IRB approach, however, banks estimate their own LGD from internal workout data, subject to regulatory floors.
Here’s why LGD is important practically: a higher LGD assumption produces a higher risk weight, which in turn produces a higher capital charge. Consequently, a bank that under-models LGD on a large unsecured portfolio ends up holding less capital than the actual loss potential warrants. This is exactly the gap regulators scrutinize during model validation and supervisory review.
IFRS 9 requires banks to provision for expected credit losses using the same PD × LGD × EAD structure. This structure applies across Stage 1, 2, and 3 assets, depending on how much credit quality has deteriorated. As a result, LGD estimates need to reflect current conditions and forward-looking scenarios, not just historical averages.
In India, the Reserve Bank of India’s Expected Credit Loss (ECL) framework, effective April 1, 2027, mirrors this approach. It shifts Indian banks away from incurred-loss provisioning toward a forward-looking model, built on internally estimated PD, LGD, and EAD. Under both frameworks, therefore, an inaccurate LGD estimate translates directly into an inaccurate provision. This either overstates earnings through under-provisioning, or unnecessarily constrains lending capacity through over-provisioning.
LGD shapes loan pricing long before any default happens. A lender pricing a loan factors in expected loss, and since LGD drives half of that calculation, it directly affects the credit spread a borrower pays.
This is why collateral matters so much in commercial lending. For example, a borrower offering strong, liquid collateral lowers the bank’s expected LGD, which typically earns a lower interest rate. A borrower with weak or no collateral, on the other hand, pushes LGD higher, and the pricing reflects that. Ultimately, risk-based pricing frameworks that ignore LGD, or rely on rough approximations, systematically mis-price credit — undercharging high-severity borrowers and overcharging low-severity ones.
PD modeling benefits from decades of relatively clean default data and a statistical distribution that behaves well. LGD, however, doesn’t have either advantage.
Recovery outcomes depend on collateral type, legal jurisdiction, the efficiency of the workout process, and the state of the economy when the default happens. Because of this, recovery data takes years to mature; workouts on defaulted loans can run for years before resolution. In addition, LGD outcomes tend to cluster near the extremes — many loans get almost fully recovered, while others get almost fully lost. This pattern breaks the assumptions behind standard linear regression.
This is exactly why LGD estimation, not PD estimation, tends to generate the most disagreement between banks, auditors, and regulators during model reviews.
Lehman Brothers’ 2008 collapse shows how badly recovery expectations can move. It also shows why LGD estimates need real stress-scenario grounding, not just long-run averages.
On the day Lehman filed for bankruptcy in September 2008, market prices on its senior bonds implied a recovery rate of roughly 30% for senior creditors. A month later, however, that implied recovery rate had collapsed to around 9%, matching the outcome of Lehman’s credit default swap auction. Anyone modeling LGD off pre-crisis assumptions, therefore, would have badly underestimated the loss severity in that window.
The story didn’t end there. Roughly two and a half years after the filing, Lehman’s estate projected eventual creditor recovery at just 16%. Over the following decade, though, as the estate liquidated assets and resolved legal claims, actual recoveries for senior unsecured creditors climbed well above that early projection, reaching close to 40% by the time of the final distributions — according to research published by the Federal Reserve Bank of New York’s Liberty Street Economics.
That range — a market-implied 9% at the depth of the crisis, versus roughly 40% a decade later — illustrates exactly why Basel’s Advanced IRB framework requires downturn LGD estimation. A model calibrated only on calm-market recovery assumptions would have missed the loss severity precisely when it mattered most, in the weeks immediately following the default.
Q1. What is the difference between LGD and PD?
PD measures the likelihood that a borrower defaults. LGD, by contrast, measures how much a lender loses if that default happens. Both feed into the Expected Loss formula, but they capture different dimensions of credit risk.
Q2. Why is loss given default important for bank capital requirements?
Basel III uses LGD as a direct input into risk-weighted assets. So a higher LGD assumption increases the capital charge on an exposure, meaning inaccurate LGD estimates lead directly to mis-calibrated capital requirements.
Q3. Does LGD matter for retail loans, or only corporate lending?
It matters for both. Retail mortgages typically carry lower LGD because of strong collateral coverage. Unsecured retail products like personal loans and credit cards, however, carry higher LGD. Basel and IFRS 9 frameworks require LGD estimation across both segments.
Q4. How often should banks update their LGD estimates?
Regulatory expectations under Basel and IFRS 9/RBI’s ECL framework call for periodic recalibration. This is especially true after economic downturns or shifts in collateral markets, rather than relying on estimates built during a single benign period.
Why is loss given default important, in the end? Because it converts the probability of a default into an actual number a bank has to plan for: in capital held, provisions booked, and prices charged. Basel III, IFRS 9, and RBI’s ECL framework all treat it as a core input, not a secondary adjustment. Understanding LGD in depth, therefore, including how to estimate it properly and where standard modeling approaches fall short, is foundational for anyone working in credit risk.
Understanding the theory behind PD, LGD, and EAD is the first step. Building bankable, interview-ready models — in Python or SAS, on real credit datasets, aligned to Basel and IFRS 9 — is what actually moves a career forward.
Explore Dexlab Analytics’ Credit Risk Modeling certification program to build PD, LGD, and EAD models from scratch, work through IFRS 9 ECL frameworks, and learn model validation techniques used by practicing risk teams.
.
Credit Risk, credit risk analysis, Credit Risk Analytics And Modeling, credit risk analytics training, Credit risk certification India, Credit risk modeling course India, Credit Risk Modelling, Credit Risk Modelling Using SAS, Credit scorecard modeling, Data and Risk Analytics, Expected credit loss ECL, IFRS 9 modeling course India, Logistic regression scorecard, PD LGD EAD modeling, Python credit risk modeling, risk management courses online, Risk Management in Banking
Comments are closed here.