A new machine learning-based corporate credit risk model showed significant improvements in predicting changes in credit ratings, according to the results of a collaboration between data and AI firm SAS, investment manager Man Group, Pension Insurance Corporation and Stanford University.
The model is intended to provide early warning indicators of credit rating upgrades and downgrades, allowing investors and portfolio managers to act before such changes are fully priced in by the market and reflected by rating agencies.
model design
The model incorporates over 20 years of historical data, including the KRIS default probability index, bond spreads, yields, equity performance indicators, and macroeconomic variables. Utilizes machine learning technology to predict the likelihood that a company’s credit rating will be upgraded, downgraded, or unchanged. Backtesting results show that this approach more accurately ranks companies by future rating changes compared to traditional risk assessment tools.
Possibility of early warning
The system is positioned as a forward-looking risk management aid, providing signals ahead of market prices. This is particularly relevant for credit rating-sensitive investors and asset managers, such as insurance companies and fund managers with regulatory investment grade requirements. More accurate predictions of rating trends are expected to inform capital allocation and portfolio management decisions.
“Our new model’s groundbreaking approach shows that investors can take action much better than current best practices. The model’s early warning signals give us critical time to act before the market has fully priced in an event, helping us better manage risk, reduce losses and seize opportunities,” said Stas Melnikov, Head of Quantitative Research and Risk Data Solutions at SAS and member of the team that developed the new model.
credit market stress
The timing of the model’s introduction comes amid rising borrowing costs and pressure on corporate credit conditions. Although the Bank of England has kept interest rates at 4%, refinancing costs are still higher than in recent years, putting pressure on businesses that need to roll over debt and increasing the risk of default. The model is designed to warn of potential rating changes before asset prices fully reflect such risks, which could impact the availability and cost of credit for a company and the valuation of existing corporate debt held by institutional investors.
Data characteristics
The study utilized more than 500,000 records covering credit events from 2001 to 2024. The study found that KRIS’ one-year probability of default (KDP) is the third most influential factor in machine learning model predictions, after option-adjusted spread (OAS) and yield to maturity (YTM). The team said the use of KDP provided information that helped identify companies at risk of rating changes prior to market pricing.
Bond ratings play a vital role in financial markets. When a company’s debt is downgraded below investment grade, policy requires some institutional investors to sell their holdings, potentially causing a decline in the prices of the affected securities. By identifying such risks early, asset managers have more flexibility to proactively adjust their portfolios.
industrial use
Insurers and asset managers are expected to benefit from the model’s insights as they are subject to regulatory and risk constraints related to credit quality. This new tool is integrated with SAS’ suite of risk management products to help you manage assets and liabilities, monitor credit risk, and prepare for expected credit losses.
Man Group President Stephen Desmiter commented on the model’s performance: “We were surprised by the results. The new model significantly outperformed the traditional approach. In layman’s terms, it was much better at telling us which companies were at risk of downgrade or upgrade.”
