Journal of Credit Risk

Incorporating small-sample defaults history in loss given default models

Aneta Ptak-Chmielewska and Paweł Kopciuszewski

  • Constructing two regressions instead of one is the correct way to estimate LGD.
  • For a low-default portfolio, applying two regressions can give results better than a single model.
  • The inclusion of small samples in the LGD estimation requires ensemble modeling techniques.

This paper proposes a methodology for estimating loss given default (LGD) that accounts for small default sample sizes. Regulatory guidelines for estimating LGD state that it is essential to take into account all observed defaults from the selected time period. In many portfolios, such as those for many emerging markets, the number of observed defaults is limited. Further, methodologies that exclude information from unresolved defaults produce biased estimates. We show that the distribution of LGD is bimodal, and we develop an estimation approach that uses a combination of logistic and linear regression to produce more reliable estimates. We implement our methodology for a sample of loans of a Polish bank, and demonstrate refinements to our modeling that produce the most precise final estimates of LGD.

Sorry, our subscription options are not loading right now

Please try again later. Get in touch with our customer services team if this issue persists.

New to View our subscription options

You need to sign in to use this feature. If you don’t have a account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here