IFRS 9 Hypothetical Portfolio Study
Banks who will or have adopted to the new IFRS 9 accounting standards for impairment face at the moment one important question:
How do our final model estimates compare to our peers? What drives the difference in those numbers?
In order to support our member banks, GCD has conducted a benchmarking study in Q4/2017, which allows participating banks to compare their final model parameters and functionality anonymously with peers. Participating banks are also able to identify the reason for those differences by tracking back to detailed components. GCD has worked with members over 6 months to design the study to take into account member needs, as always.
- The study results have been discussed with members and regulators and shared with the financial community on various occasions. The study indicates that IFRS 9 credit loss estimates vary widely.
- It is the first study worldwide which quantifies the differences in the methodologies the banks used to calculate their credit loss estimates. The variability factor is calculated by taking the highest ECL per borrower provided by a bank and dividing it by the lowest value provided by a bank.
- The variability factor varies by asset class, country, facility type and creditworthiness (PD) of the hypothetical borrower. For Large Corporates - on average - the variability factor for the 12-month ECL was 12 if banks have used a common macroeconomic scenario. In other words, one bank estimated a lifetime expected credit loss (ECL) that is 12 times higher than another bank’s ECL estimate for the same hypothetical borrower. Under IFRS 9 and CECL, banks are required to use their own macroeconomic forecast, which further increases the variability between banks.
- The main drivers for this variability between banks lie in the different methodologies, data sources and assumptions used to derive point-in-time probabilities of default (PiT PDs), loss given default (LGD), multi-year PD curves, and expected life-time (maturity) for revolving facilities. These differences occurred even though participating banks based their estimates on a common macro-economic forecast and a hypothetical portfolio and were provided with detailed specifications (e.g. a given maturity, a fixed loan-to-value (LTV) ratio, a pre-determined industry) for each hypothetical borrower to be assessed.
Find out more:
For more information please consult our publication in the RMA journal. Members have also access to more detailed results - please log in at our member website in order to be able to download the report.
Participants to the study have received a much more detailed peer comparison report and a complete data return.
The study will be re-run in 2018 with the following timelines:
- April 2018 to Mid May 2018: Discussion with member banks on further improving the templates
- End May 2018: Templates to be distributed to participating banks
- July 15th, 2018: Final timeline to deliver the data templates through GCD's data portal
- August 2018: Report creation and data return
The data templates of the 2018 study are available here (log-in of members required).
Also non-members can join the study. If you have any questions, or if you are interested in participating, please contact us for further information.
Note: in 2018 we will also test the hypothetical portfolio under stress conditions, as being defined by the EBA in their Stresstesting Excercise 2018. EBA-regulated banks have therefore the possibility to re-use the work they have developed for the EBA stresstest.
Why should your bank participate:
- Determine how your bank's IFRS 9 estimates compare to that of peer banks.
- Neutral to your bank's portfolio or macro-economic forecast
- Be able to track down the reason for the variability, e.g.
- Does the difference lie in the PD estimation, the LGD estimation or the exposure calculation?
- How much is the 12-month ECL or the lifetime ECL impacted by banks' economic forecast?
- Is your bank's stage allocation process more or less conservative than that of other banks?
- Do banks PD curves differ by country? How many banks apply LGD term structures?
- In which countries is the variability between banks the most?
- How does the expected life of revolving facilities differ between banks?
- "By banks, for banks": We value your input in changing the study to your needs
- Aligned with GCD's other datapools: Easily explain further differences making use of GCD's
- Benchmarking platform: Benchmarking estimated PD, LGD and CCFs on name-by-name basis / by “risk cluster”
- Ratings and PD platform: Benchmarking observed default rates by “risk cluster”
- LGD/EAD platform: Benchmarking observed LGDs and CCFs
- Based on GCD's datapooling infrastructure: Highly-secured data portal and datapooling regulations ensure maximum confidentiality
- Ultimately, strengthen your IFRS 9 model development and validation process