LoanBook Severity Model v3

The Severity Model calculates loss given default (LGD): the percentage of a defaulted loan balance, measured at the time of default, that is lost after liquidation and recovery. LGD combined with CDR produces expected annual credit losses. Severity varies by loan type, collateral, geography, and market conditions.

Methodology

Severity v3 is the production model on the sectors where it has been calibrated; the v2 component engine remains underneath it for the rest.

Loss distribution: Severity is modeled as a Beta distribution rather than a single point, so stress scenarios draw from the tail of the same distribution the base case uses.

Joint macro frailty: Default and severity share a macro factor, so severity rises in the same environments in which defaults rise.

Cause-specific severity: Losses are estimated separately by the cause of liquidation rather than from one blended average.

Residential: Severity is read from a realized loss table built from Fannie Mae loan performance data. The property is marked to its value at the time of default, in the kind of market loans default in, and a second lien takes the loss on the combined debt first.

Missing inputs: A missing LTV or FICO makes no adjustment. The model never substitutes an assumed value.

Government guaranteed portions carry no severity: the government pays the investor at par. Lease severity uses the lease program's own loss assumption.