The Biocomplexity Institute is an academic partner of the Chickasaw Nation—a federally recognized sovereign Indian nation spanning 13 counties in Oklahoma. This collaboration supports the Chickasaw Nation’s Office of Investment Management (CNIMO), which oversees the Nation’s investment funds. Our primary objective as an academic partner is to enhance their forecasting and investment modeling capabilities, with a particular focus on asset price prediction and portfolio management strategies.
As part of the project, the Biocomplexity Institute developed models to forecast yields on U.S. Treasury securities, a traditionally low-risk asset. The team designed and evaluated a range of statistical and machine learning models to predict 12-month Treasury yields for securities with maturities ranging from 1 to 10 years. Many of these models incorporated key economic indicators to enhance forecast accuracy. To improve robustness, individual model forecasts were combined using various ensembling techniques. The performance of these models was benchmarked against the no-change-in-yield model, the standard baseline in yield forecasting.
FINDINGS: Our models outperformed the baseline no-change-in-yield model in over 60% of cases when evaluated across a 24-year period (2000–2024).