What we do
In 2009 Global Credit Data started a study of Observed Default Frequencies, i.e. the number of defaults actually counted by a bank in a year within a segment of its obligors. This exercise complements the analyses on LGDs and EADs, and, like them, is strictly historical. In 2015, the data template and process has been amended in such a way that also migration matrices and multi-year defaults rates (both relevant for stresstesting and IFRS 9 impairment modelling) are calculated and given back to the participating banks.
The database input for this datapooling is much simpler than the LGD-EAD database: we are simply asking for quarterly “portfolio snapshots”, including basic information such as asset class, country/geographic region, industry, rating category and PD. All this enables us to calculate and give back default rates, migration rates and average PDs for a vast amount of different segments – information which member banks can use instantly to benchmark their internal PD systems.
How can the database be embedded in your regular processes
Since our last collection exercise our database contains more than 1.5 million obligors and 84,000 defaults through the last economic cycle (from 2002 until 2015) from 24 banks, split out by year, rating, region, industry and asset class. This database allows us peer-benchmarking of internal rating models, construction of correlation matrices and long-term default rates.
Some concrete use cases are:
- Benchmark your PD masterscale (comparison PDs and default rates per rating class)
- Benchmark your system’s discriminatory power
- Identify macro-economic dependencies in default and migration data: Extract a “systemic factor” from rating migrations or default rates
- Benchmark your asset correlations and long term default rates
- Benchmark your stage allocation / SICR buckets (thresholds for “life-time PD” movements) under IFRS 9
- Reduce uncertainty add-ons for lack of data