Paper: arXiv 2511.11364
Authors: Pomazanov Mikhail
Abstract
The paper shows how to determine the loss on an LGD borrower’s loan after default, with or without preparation of a separate model. LGD after default is estimated taking into account the average repayment period of the defaulted loan, knowledge of volumes, moments of default and repayments, the rate or other parameters in the vector of determinants. The calculation of the average repayment period for overdue loans is given in the article. A Bayesian scheme is used to estimate repayable debts, considering the percentage of repayment. A general recovery model was used for the LGD segment recovery process. Only this type of model allows you to set LGD less than or equal to 1, which is required for further estimates.
Complexity vs Empirical Score
- Math Complexity: 6.0/10
- Empirical Rigor: 5.5/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper presents a structured mathematical framework for modeling Loss Given Default in default (LGD ID) using continuous discounting, proportional collection cost allocation, and exponential recovery dynamics. While it establishes useful mathematical bounds (LGD <= 1) and optimization schemes for parameters, the empirical validation remains largely synthetic or relies on summarized portfolio statistics rather than extensive backtesting on open datasets.
Research Flowchart
flowchart TD
A["Research Goal: Estimate Loss Given Default (LGD) After Default"] --> B["Inputs: Repayments, Default Moments, Volumes & Vectors"]
B --> C["Calculate Average Repayment Period"]
C --> D["Apply Bayesian Scheme for Repayable Debts"]
D --> E["Run General Recovery Model"]
E --> F["Outcome: Precise LGD Post-Default (Ensures LGD ≤ 1)"]