Model Development and Model Validation: Creating a Sound Cross-Divisional Relationship

SAN FRANCISCO - Oct. 15, 2014 - PRLog -- Model risk issues are as important as ever in the environment where the regulatory pressures are mounting and resources are scarce. Today, model development, validation and assessment require new tools and a range of quantitative and qualitative measures to tackle model risk issues. At the 5^th Edition Model Risk Conference (http://www.global-fmi.com/marcusevans-conferences-event-details.asp?EventID=21608&SectorID=2&utm_source=YLintvw&utm_medium=link&utm_campaign=21608_YLintvw_link), Yidong Liu with MUFG Union Bank, N.A. will touch upon how to facilitate positive communication between model developers and validators in his stand alone case study presentation. Yidong gives pre-conference insights into his session and the conference topics:

What are the key trends and developments in model risk management today?

YL:Currently, model risk management involves regulators, an audit department, a model validation department, a model development department, and model users. These groups are not clear where the boundaries of their duties end. Because of the lack of clarity, communication between each group is critical at present.

To ease the tension between these groups, a well accepted protocol is necessary. One that is understood by all and calls for cooperation across the multiple company divisions.

At the conference, you will be talking about relationships between model developers and model validators.  What are the major challenges in these communications?

YL: Model validators and model developers have common interests, but different goals. The common interest is that we all work for the same bank. The goals of model developers include: satisfying model users, using available resources to address business issues, passing model validation and regulatory requirements. The goals of model validators include: following the model validation procedures and satisfying auditing and regulatory requirements.

The main challenge is that the same model validation procedure cannot be applied to each model since they are different. Benchmarks testing and back testing are not available for economic capital calculation. In addition, scenario testing is hard to apply to CCAR models since the correlation under stressful scenarios is usually not the same as the correlation among normal scenarios.

How can relationships improve and become stronger between model developers and validators?

YL: First of all, model developers and model validators should respect each other and understand the limitation from each side.

Model validators need to be aware of the data and people resource limitations that developers face in their operations. On the other side, model developers need to take into account the knowledge and time limitations that hinder project efficiency for model validators.

To fix this void between the two, companies need to train model developers and validators together.

What is the role of internal audit in model development and validation?

YL: Because the boundary between model validation and audit is not clear, it is important for the auditor and model validator to work closely together.

One solution is to perform auditing and model validation at the same time. In this system, model developers save time by avoiding explaining the same topics twice. This approach also avoids any gap between auditing and model validation. I have personal experience in checking system implementation data quality from both perspectives (model validators and auditors).

What do you think you will gain by attending this event?

YL: I look forward to learning from each other. If other banks have had similar experiences and found effective solutions, then we will save time compared to trying to find solutions on our own.

Although different banks have different model risk management procedures, they all have similar models. Understanding the limitations and assumptions of common models will help people understand real model risks and find effective ways to measure/manage model risks.

All responses represent the view of Mr. Liu and not necessarily those of MUFG Union Bank, N.A.

For more information, please contact Tyler Kelch, Marketing Coordinator, Media & PR, GFMI at 312-894-6310 or tylerke@global-fmi.com

About Yidong Liu

Director, Commercial Credit Scorecard Modeling, Credit Strategies Group, MUFG Union Bank, N.A.

Yidong Liu is a financial risk veteran with 18 years of working experience in different financial institutes. Yidong started his first job as a model validator at Global Analytics in CIBC bank at Toronto, Canada. At that time the model validation was called vetting. The team validated models in trading floor systems and models implemented in Excel. Later on, Yidong worked on model validation at PNC bank and BBVA Compass Bank.

Currently, Yidong is the manager of the Union Bank wholesale scorecard PD modeling team. The team supports scorecard modeling, CCAR stress testing and acquired asset valuation. Yidong Liu received his MS in Computer Science from University of Toronto and his MS in Applied Math from University of British Columbia. Yidong also holds CFA and FRM certificates.

About GFMI

GFMI (a marcus evans company) conferences annually produce over 2,000 high quality events designed to provide key strategic business information, best practice and networking opportunities for senior industry decision-makers. Our global reach is utilized to attract over 30,000 speakers annually; ensuring niche focused subject matter presented directly by practitioners and a diversity of information to assist our clients in adopting best practice in all business disciplines.