By Xiaojing Li, Director, Quantitative Methods, CoStar Group.
Xiaojing, can you please tell the Risk Insights readers a little bit about yourself, your experiences and what your current professional focus is?
Xiaojing Li is director of Quantitative Methods of CoStar Portfolio Strategy and CoStar Risk Analytics (consulting subsidiaries of CoStar Group). She is currently in charge of model development and validation of CoStar’s core mortgage risk service product/CRE credit risk model – CompassCRE and CompassCMBS. She leads the quant team on building various models and analytical tools for analyzing term default risk, maturity default risk, prepayment risk, and loss severity, assessing risk-weighted asset and capital reserve requirement under Basel III compliance, and setting loss reserves under FASB’s CECL credit loss accounting standard. Ms. Li has almost a decade of experience working with BHC clients on producing custom scenarios and regulatory capital adequacy analysis in CCAR/DFAST framework.
Before joining CoStar Group, Ms. Li was a research assistant in the field of Consumer Finance at Ohio State University (OSU), where she majored in econometrics and consumer finance and received her M.S. and Ph.D. degree in Economics.
At the CECL Congress 2018, you will be speaking on your insight regarding, ‘Developing top down and bottom up approaches for an efficient and accurate CECL solution’. Why is this a key talking point in the industry right now?
To achieve the goal of creating “reasonable and supportable” forward-looking view, which is one of the major requirements under CECL and critical step for any organization implementing CECL, the two most important components in the modeling/analytics system, in my opinion, are sound and rigorous forecast methodology from top down perspective and the sufficient and granular data from bottom up perspective. Lacking either of the two components would pose significant risk of failing the goal of creating “reasonable and supportable” forecast.
What are the key considerations that need to be made when identifying macro-economic variables relevant to the specific industry?
For a specific industry like commercial real estate, key considerations made when identifying macro-economic variables include finding and tracking the dynamic relationship among broader economy, capital markets, and CRE space markets and CRE asset markets. The relationship needs to be investigated with sufficient time series as well as cross-sectional data, and supported by statistical significance and reasonable economic implications.
The common primary drivers within broader economy, capital markets and monetary policy need to be identified to establish their relationship with CRE market performance from long term and aggregation level perspective. Specific macro-economic variables that may impact the sub-sectors within CRE markets differently also need to be considered.
When building econometric models using time series data, certain tests need to be conducted first, such as stationary test, to ensure the soundness of the regression result.
Why is it an important to develop multiple baseline scenarios?
The forward-looking view of any industry is formed based on the view of inputs, which in turn are created using predictions and estimations. As such, all forecast is hypothetical and subjective, and to a bigger extent when the forecasting horizon is extended longer.
Developing multiple baseline scenarios can provide a wider range of possible outcomes and thus add flexibility in the later usage of the forecast. More in-depth sensitivity and impact analysis can be conducted based on multiple baseline scenarios using different macro-economic or sector inputs, which could provide insight for the final decision-making.
Can you provide a brief overview of documentation and back testing?
Model documentation contains (and is not limited to) model design, methodology, implementation, input and output interpretation, model validation, model limitations, and model governance and monitor system. A comprehensive model documentation is critical to an effective model risk management.
Back-testing basically assesses the model performance by comparing the actual observations with the model outcomes by setting the model’s prediction starting period back to a historical time point. Back-testing is an important component of model validation and a key component to the model-based risk measurement and management. A comprehensive back-testing should include comparison between actual and modeled results from multiple perspective and in granular level, comparison on both magnitude and ranking power, to provide a complete overview of the model performance.
Finally, what challenges do you foresee with CECL implementation over the coming years and how can institutions best plan to meet deadlines?
In my opinion, the challenges are to seek the common ground of understanding on the new concepts and requirements, and then identify potential weaknesses or the insufficient areas that need more data, more efficient analytic tools, or a better system. Better knowledge sharing and experiencing borrowing, together with proactive exploring third-party data sources could increase the transparency and efficiency for all institutions in general when implementing CECL.