New risk model sets benchmark for accuracy in predicting credit rating changes

A new study has unveiled one of the most accurate corporate credit risk forecasting models to date. This study is a result of collaboration among quantitative finance experts from data and AI leader SAS, investment management firm Man Group plc, UK insurer Pension Insurance Corporation plc (PIC) and Stanford University. It provides an early-warning indicator that flags potential rating changes before they are priced into the market and credit rating agencies act.

This machine learning model leverages multiple features, including over two decades of KRIS^® (SAS^® Kamakura Risk Information Services) default probability data. The model was significantly better at ranking firms by predicted probability of credit rating change – upgrade, downgrade or no change – than traditional approaches using legacy tools.

“The breakthrough approach of our new model shows that investors can do a lot better than the current best practices,” said Stas Melnikov, Head of Quantitative Research and Risk Data Solutions at SAS and a member of the team that developed the new model. “The model’s early-warning signals give them critical time to act before the market fully prices in the event, helping better manage risk, reduce losses and seize opportunities.”

With its unprecedented accuracy in identifying firms with elevated probabilities of downgrade or upgrade, the new model facilitates proactive portfolio management and improved capital allocation. Its insights are particularly important for investors who are highly sensitive to the credit ratings of securities on their balance sheet. For example, the ability to more consistently and accurately forecast rating transitions will enhance risk management and credit selection for insurance companies.

Even after the US Federal Reserve’s most recent rate cut, today’s high interest rates are squeezing corporate borrowers. Refinancing costs are climbing as corporate credit profiles weaken. This raises the risk of defaults, which affect investors, including money market and hedge funds as well as banks and insurance companies that hold corporate bonds.

The model can be used as an early-warning system, proactively identifying potential ratings changes before that risk is fully reflected in the prices of the securities. It is based on AI and machine learning techniques and KRIS, which provides comprehensive and timely assessments of credit risk for public companies. Among other data, KRIS includes a one-year default probability (KDP), which is based on a multifactor reduced-form credit model incorporating macroeconomic, market and financial data.

The model, further explored in this SAS Voices blog post by Melnikov, provides one of the most accurate, realistic forecasts of corporate-bond rating migrations to date.

“The results surprised even us,” said Steven Desmyter, President of Man Group – a SAS customer – in a recent LinkedIn post about this research. “The new models significantly outperformed traditional approaches. In plain English, they were much better at telling us which companies were at risk of downgrade or upgrade.”

The team worked with data that included more than a half-million observations covering two decades of credit history (2001-2024). The data comprised bond spreads, yields, credit default probabilities, equity momentum and macroeconomic variables.

Illustrating the value of KRIS data, the study showed that KDP was the third-most influential feature in the machine learning (gradient boosting) model, behind only option-adjusted spread (OAS) and yield-to-maturity (YTM). In other words, through KDP, KRIS added additional orthogonal information that helped identify opportunities not yet priced into the market.

What does this mean? If a portfolio manager or other investor can predict a credit downgrade before it happens and before the market prices it into a company’s valuation, they can achieve better returns and/or lower their risk.

Rating agencies assign a credit rating to each company that is a measure of that company’s creditworthiness. Any changes in that rating affect market perception of creditworthiness and can lead to a repricing of its debt.

Additionally, many institutional investors are mandated to hold only investment-grade debt. If a company’s debt (e.g., bonds) is downgraded from investment grade to high yield, for example, certain investors must disinvest, which can drive down the market price of that debt.

SAS acquired Kamakura Corp. in 2022 to enhance its financial risk management offerings, particularly in asset and liability management (ALM). Today KRIS and other SAS offerings are part of a comprehensive and integrated suite of risk management solutions for the financial services industry. These include solutions for:

• ALM.
• Credit risk management.
• Enterprise stress testing.
• Expected credit loss.
• Insurance risk management.
• Risk governance.

Learn more at sas.com/risk.

SAS is a global leader in data and AI. With SAS software and industry-specific solutions, organizations transform data into trusted decisions. SAS gives you THE POWER TO KNOW^®.