Artificial Intelligence Helps Fixed Income Investors Predict Credit Rating Changes
When it comes to data, fixed income investors would tell you the more the better. Until it comes to processing and analyzing that data, in order to make decisions that could be worth millions of dollars.
“In fixed income, you cannot be wrong given the amounts involved. The dollar amounts are much more material, and the losses can be too.” said Don Povilaitis, Director of Credit Research, Fixed Income.
AIMCo’s Fixed Income team is getting a boost in its efforts from a machine learning and artificial intelligence (AI) model that’s helping to predict credit ratings. Since the Great Recession, credit rating agencies like Standard & Poor (S&P) and Moody’s — the main rating agencies — have become more erratic in their assignment of upgrades and downgrades in ratings. This volatility is difficult for humans to track and predict and can lead to bigger swings in corporate and government bond returns.
Now, with modern computing power, and with advanced machine learning and AI, predicting credit ratings can be much more efficient to help drive returns for AIMCo’s clients. The AI model draws on data daily and is trained to predict rating changes. It learns as more data gets processed. Portfolio managers scrutinize the model’s output as it consistently retrains itself to enhance results. The process involves millions of computations which would have not been possible a few years ago. The model provides both a prediction of credit directionality and certainty and has kept on improving since implementation at the end of last year.
“AIMCo is ahead of the curve. As more data gets fed into the model, AI provides a competitive advantage to portfolio managers to pre-emptively take advantage of future upgrades or downgrades which leads to robust returns for the integrity of the process for our clients,” explained Povilaitis.
The team using the model has been quite impressed with its accuracy. Recently, the model predicted an upgrade in U.S. homebuilders which was flagged early in the COVID-19 pandemic.
The model is a product of AlphaLayer, AIMCo’s joint venture with AltaML that puts machine learning and AI to work in investment management. The partnership was developed in 2019 to help drive operational efficiencies, manage risk and create an alternate source of revenue for AIMCo’s clients.
“The success of this model was not only in its ability to provide accurate credit ratings predictions but that it’s fully operational and integrated into the workflow of the Fixed Income team,” said Chad Langager, General Manager of AlphaLayer. “This is a big win for AlphaLayer on its mission of delivering applied AI solutions that generate ROI for AIMCo.”
Langager expects that the existing model has near-term applications within public equities. In the future, it could be expanded to a broader universe of securities outside of North America and improved in its accuracy through the addition of alternative data, such as Environmental, Social and Governance (ESG) data.