Low Default Portfolios (LDP)

Banks face significant challenges in building internal probability of default (PD) models for low default portfolios (LDPs) such as large corporates and global banks. In particular, scarcity of internal default data for these portfolios typically leads to inaccurate PD models with low predictive power.

Hence, banks often resort to external datasets, such as third-party default data and external ratings, to build their internal PD models. Such methodologies often result in conservative PD estimates, which in turn lead to high regulatory capital charges.

RISE's LDP solution addresses this industry-wide problem by pooling internal default data across banks. Our solution therefore facilitates the development of more accurate and robust PD models for LDP. Currently, our database consist of –

2 portfolios 'Large Corporates' and 'Global Banks'

90,000+ obligor
level data

70+ countries
covered

Value Drivers

Rich data pool
Continual expansion and update of data

Suite of modelling methodologies
Centralised modelling infrastructure

Comprehensive model documentation
Ongoing model monitoring

Key Benefits

Avoid regulatory capital add-ons

Optimised regulatory capital by using better PD models
Reduced capital overlays and buffers

Revenue enhancements

Improved transaction pricing
Better risk-return trade-offs

Cost savings

Efficiencies by avoiding potential duplication
Accelerated model redevelopment

FAQs

Which portfolios are covered by the LDP solution and why?

The large corporates (revenues >EUR 500 million) and banks portfolios are covered, as they are both:

1. Material portfolios for most banks; and,

2. Portfolios where scarcity of defaults is most pronounced

What data inputs will banks need to provide for the LDP solution?

Banks will only need to provide obligor names, common attributes such as domicile and industry segment, and default attributes such as status of default and default date on an annual basis. Sensitive data such as obligor financial data and exposure data will not be sought

How will data confidentiality be ensured?

We have incorporated an anonymisation process which effectively removes all references that map any data to the contributor bank

Can banks just subscribe to the LDP solution to receive pooled data, while avoiding data sharing with RISE?

No, because we use a 'give-to-get' model. Only banks that contribute data can receive the pooled data in return

What are the expected use cases for the banks?

Banks have the option to leverage the LDP solution in three ways. They can use the:

1. Default data pool to develop their own models in-house;

2. RISE PD models as challenger models; or,

3. RISE PD models as official champion models

Are the RISE PD Models transparent to users or will they be a black box?

Comprehensive model documentation is provided by RISE as part of the PD models, which provides complete transparency into all the modelling methodologies used