Member-only story
Data ScienceDataData Analysis
How to validate LGD data in 3 steps
Three checks that catch what your LGD model is hiding
Biswarupa Mohapatra
3 min read7 hours ago
Press enter or click to view image in full size Photo by Infrarate.com on Unsplash
A bank reported an average LGD(Loss given default) of 25% on its corporate loans.
Then someone counted the defaults. 15 in the books. Only 12 in the data.
The 3 missing ones? The biggest losses.
The first time, I only checked the defaults list and skipped the accounting reconciliation.
Step 1: Count the defaults
Are all of them in your data?
Take the list of defaulted companies from the credit or watchlist team and compare it with the LGD dataset. Missing cases are usually the worst ones, and leaving them out makes LGD look better than it is.
Example: The bank had 15 defaults, but the dataset had only 12.…