Outlining a CCAR strategy beyond model development
1. Outlining a CCAR strategy beyond Model Development
Much of the discussion around CCAR has been focused around scenario
modeling. While i agree that scenarios are a major part of CCAR, they only
represent one half of the picture. The other half of course is related to sourcing
quality Data.
Scenario modeling is a complex process, which requires banks to assemble
teams of consultants, data scientists, PHDs, and economists to debate and
develop various aspects of the regulatory requirements. While this is not an easy
task, most organizations are able to quickly build or buy the necessary resources
to accomplish it.
On the other hand,sourcing high quality Data is more than an intellectual
exercise;it requires numerous stakeholders to work together, and in many cases
ends up being the larger challenge for CCAR institutions.
The equation for CCAR is as follows:
Quality Data + Quality Models = Accurate and Scenario modeling
So how does one source quality data? This question is often asked, and rarely
has an easy answer. A bank holding company’s (BHC) ability to source high
quality data is dependent on sound Data Governance practices. Most CCAR
banks have spent the past two decades defining and improving Critical Business
Processes (CBPs) but have spent little time identifying Critical Data Elements
(CDEs).
As new banks are brought into the CCAR fold, and as regulators become more
intelligent around Data Governance (DG), most banks are scrambling to
establish robust DG programs. But unlike models, the effort required to
implement DG programs and to manage CDEs,is many fold larger, more
complex, and requires massive organizational collaboration.
Upto this point, defining Critical Data Elements (CDEs) and creating governance
mechanisms has been an after thought, if not completely neglected by some
banks.
Unlike modeling, throwing resources at the DG program can have diminishing
returns, especially if the organizational support is not explicitly in place. Most
organizational data is siloed and as such falls victim to the standard territorial
conflicts. How many of us have sat thru meetings and have heard both business
and IT teams refer to such structured data as “My Data”.
2. To implement effective Data Governance, Data Quality and CDE lifecycle
management, there are four main work streams that need to be managed
concurrently (I will cover each of these streams in a separate post at a later date)
1. Definition and approval of company wide data policies e.g. Data
Ownership, Definition of Critical Data Elements and Data Quality etc.
2. Identification of Critical Data elements needed for CCAR reporting
3. Adoption of standardized messaging standards e.g. ISO 20022
4. Tools e.g. Business Glossary, Data Profiling and Lineage, Quality etc.
As long as Data Governance and Data Quality are considered an after thought
instead of precursors, the CCAR programs will continue to deliver abysmal
results and regulators will continue to hand out MRAs.