Public Sources and Limits for Credit LGD and PD Data

Summary

The document discusses public data sources for comparing loss given default (LGD) and probability of default (PD) across banks and other financial institutions without relying only on a Merton model. For LGD, it points to credit event auction settlements and explains that recovery outcomes can vary by debt seniority. In some bank cases, subordinated debt had no deliverable obligations following bail-ins, producing settlements at par and making auction-derived recovery estimates unrepresentative of ordinary claim losses.

The answer treats auction results as a demonstrable public source, while warning that actual bilateral claims may differ because counterparties can agree to case-specific settlements. For PD data, it suggests ratings agencies but provides no particular dataset or link. The response therefore offers a starting point for data collection, not a comprehensive or directly comparable bank credit dataset; users would need to consider instrument type, seniority, and the gap between auction settlement recoveries and actual exposures.

Key ideas

• Credit event auction settlements can provide public observations for estimating recovery and LGD.
• Auction recoveries depend on the seniority and type of debt involved.
• Bail-ins and the absence of deliverable obligations can make subordinated debt auction outcomes unusual.
• Bilateral settlements may cause actual claim losses to differ from auction-based estimates.
• The answer points to ratings agencies as a possible source for PD data but identifies no specific dataset.

Tags

# LGD/PD Databases


# LGD/PD Databases












I am trying to compare LGD/PD for Banks and other financial institutions using different approach thank Merton. Are there any publicly available data on which I can build?

Thanks.

## Answer by Mehness (score 2, accepted)

https://quant.stackexchange.com/a/31341

Re LGD: you can look at Mark-iT for ISDA Credit Event Auction settlements, here's a link actually http://www.creditfixings.com/CreditEventAuctions/fixings.jsp

Obviously these will give recoveries as determined by bond auctions, varying according to SUB and SENIOR in the type of names you are looking at. These outcomes are a bit pathological for banks as SUB credit events sometimes had no deliverables given bail-ins, prior to the 2014 update to ISDA definitions, therefore $LGD = 1-R$ settled at 100% for SUB in several cases eg BES.

That may all be derivatives mumble swerve, clearly LGD on actual claims may vary from counterparty to counterparty where bilateral commutations maye have been agreed on a case by case basis (eg AIG), but it is a good demonstrable public source.

Re PDs I guess, (sadly) you have to go to ratings agencies but I don't have a link to hand.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.