Loss Given Default (LGD)

Loss given default (LGD) is the proportion of an exposure that a lender expects to lose if a borrower defaults, after counting everything it recovers through collateral, guarantees, asset sales and negotiation, and after the costs of recovery. It is usually stated as a percentage of the exposure at default. An LGD of 35% means the lender expects to get back 65 cents of each dollar outstanding once the dust has settled.

LGD is the severity half of credit risk measurement. Probability of default estimates how likely a default is; LGD estimates how bad it would be. Together with exposure at default they produce expected credit loss for accounting and expected loss for regulatory capital. The complement of LGD is the recovery rate, a term more common among bond investors and rating agencies.

Suppose you lend a friend money to buy a used car and keep the spare key as security. If the friend stops paying, you can sell the car. What you lose is the part of the loan the sale does not cover, plus what you spent on adverts and towing.

That leftover loss, as a share of what was owed, is loss given default. Lend against a car that holds its value and the loss may be small; lend with nothing to sell and you may lose almost everything. Banks estimate this share for every type of loan before they lend, because it tells them how much a default would really cost.

LGD is the complement of the recovery rate:

When it is measured from the actual workout of a defaulted loan (workout LGD), recoveries and costs are discounted back to the date of default:

• $\text{EAD}$ is the exposure at default, the amount owed when default occurred.
• $\text{PV of recoveries}$ is the present value of cash collected after default, from collateral sales, guarantors or the borrower.
• $\text{PV of recovery costs}$ is the present value of legal, administrative and selling costs.

A bank is owed $500,000 when a borrower defaults. Over two years it recovers cash with a present value of $330,000 and spends $30,000 (present value) on lawyers and a receiver. Its LGD is $(500{,}000 - 330{,}000 + 30{,}000) \div 500{,}000 = 0.40$, or 40%. Market LGD, by contrast, is read from the price at which defaulted bonds or loans trade shortly after default.

Recoveries vary far more than many lenders expect. The main drivers are:

• Seniority. Claims higher in the debt seniority ranking are paid first, so senior secured debt usually has much lower LGD than subordinated debt.
• Collateral type and cover. Cash and government bonds recover well; specialised machinery or stock-in-trade much less so. A low loan-to-value ratio means a larger cushion.
• Legal environment. Slow courts and debtor-friendly insolvency laws stretch out recovery and raise costs, which raises LGD.
• Industry and asset liquidity. Asset-light businesses such as services firms leave less to sell than utilities or real-estate owners.
• The economic cycle. Collateral values fall in recessions, so losses are higher when defaults are also higher.

Retail products differ too. Unsecured lending such as credit cards often shows LGDs of 70% or more, while prime residential mortgages with moderate LTVs can show single-digit LGDs.

Because recoveries fall when defaults rise, regulators require banks that model their own LGD under the advanced internal ratings-based approach to use a downturn LGD that reflects stressed conditions. Under the foundation approach banks use supervisory values set in the Basel framework. Following the 2017 Basel III reforms, senior unsecured claims on non-financial corporates carry an LGD of 40%, senior unsecured claims on banks and other financial institutions 45%, and subordinated claims 75%, with lower values for eligible collateral. IFRS 9 accounting, on the other hand, uses unbiased, forward-looking LGD rather than a downturn figure.

LGD data are thin and slow to collect, since a single workout can take years to close, and the most recent defaults are still unresolved when models are calibrated. Results depend on the discount rate chosen and on how cured loans (those that return to performing) are treated. Distributions are often bimodal, with many loans recovering almost fully and many losing almost everything, so an average can describe very few actual cases.

Despite these weaknesses, LGD estimates are used well beyond capital rules. They set the reserve part of a loan loss provision, shape how much collateral a lender demands before approving a deal, and help price protection bought through a credit default swap, where the protection seller in effect pays the buyer the LGD on the reference debt.

• Basel Committee on Banking Supervision