Post-webinar Q&A: Speed and Scalability – Automation in Credit Risk Modeling

This article is written and published by S&P Global Market Intelligence, a division independent from S&P Global Ratings. Lowercase nomenclature is used to differentiate S&P Global Market Intelligence credit scores from the credit ratings issued by S&P Global Ratings.

In our recent webinar, Speed and Scalability: Automation in Credit Risk Modeling, RatingsXpress® product experts from S&P Global Market Intelligence shared their top insights on the role automation plays in credit risk management, how we incorporate client feedback into our product design, and how our products can make it easier for machine learning algorithms to perform human tasks at a similar level of accuracy as humans with the speed and scalability automation brings. 

Continue reading for answers to the top questions asked during the webinar and answered by our speakers: Michelle Cheong, Director, Senior Lead of Product Development, Shruthi Nagarajan, Product Manager, and Sowmya Karre, Product Manager. 

Q: Are leverage loan prices available in RatingsXpress?

A: While leverage loan prices are not available in RatingsXpress, we have other content sets within Market Intelligence that can handle leverage loans. Within RatingsXpress, we have ratings, ratings benchmark data, as well as the default and rating transitions data and recovery data for different loan types. 

Q: What is the value of using credit ratings reports (since credit ratings tend to be slower than other market indicators)?

A: There are many potential use cases, including:

1. The early warning indicators from the text combined with the Credit Highlights or the Outlook may or may not be reflected in the market prices immediately. For example, if you look at COVID-19 reports, in January and February 2020, when COVID was relatively mild in the United States but retail was already in a debilitated state. The outlook sections provide a forward-looking view were indicating something that’s pretty stable; however, you can see that even in the early COVID cycle, as some of the reports started to be released, the text started to turn negative and the trend became more persistent and started reinforcing itself across peer companies.

2. One could look at the qualitative factors like sentiment to make sense of the data. We have Business Risk and Financial Risk scores as qualitative metrics with additional commentary for background and make sense of the credit ratings.

3. Text analysis helps discern and rank two similarly performing peer companies with the same credit rating in addition to quantitative scores. 

Q: What is aggregator sentiment text-based information and how does it work with numerical information?

A: In a report, you have a Business Risk and a Financial Risk section. In parallel, there is a business risk score and a financial risk score which is numeric information that ranks different entities according to how they fare on different level of risk resilience of something. The text complements this because it provides more context as to what the business risk drivers are. Two entities acrossdifferent borders can have the same business risk score, but from a risk management point of view, one could have a high business risk due to local government, and the other due to a macroeconomic growth driver, for example. Many will use these sections to get the context behind the numerical information – they work hand in hand. 

Q: Can the ability to ingest the data be controlled by the user?

A: Absolutely, that’s exactly why we put this data in text. We considered client feedback and put this in individual fields in the database. You can ingest that in rather than taking the entire body of text. We provide both options, but this also helps the client to control what they want to digest and ingest into the database. 

We have millions of research articles published by S&P Global Ratings. If you were to just data dump out everything into your own database, it could take a lot of time to ingest the data before you can start to work on it. So, we’ve been using the metadata to filter out only the articles that you need. This gives you the option of taking the full text, the bodies of text, and only the articles required for your analysis.