Fintech Thrust Public Seminar | Deep Learning for MBS Prepayments
Deep Learning for MBS Prepayments
Abstract: Predictions of prepayment speeds are mission-critical for investors, dealers, originators and other participants in the $10T Agency MBS market. We develop deep learning systems for prepayment speed prediction that set new accuracy standards, delivering performance-boosting edges to market participants. Our systems harness data of unprecedented size and granularity, covering monthly records for tens of millions of borrowers across the US over two decades. By uncovering hidden nonlinear patterns in borrower behavior at the individual loan level, they improve prediction accuracy for MBS pool CPRs by a full order of magnitude relative to the market’s current “gold standard.” Our predictions are robust in all market environments including the pandemic. Rigorous significance tests offer deep insights into the variables influencing predictions.
Kay Giesecke is Professor of Management Science & Engineering at Stanford University. He is the Director of the Advanced Financial Technologies Laboratory and the Director of the Mathematical and Computational Finance Program. Kay is a member of the Institute for Computational and Mathematical Engineering. He serves on the Governing Board and Scientific Advisory Board of the Consortium for Data Analytics in Risk. He is a member of the Council of the Bachelier Finance Society. Kay is also the founder, Chairman and Chief Scientist of Infima Technologies, a venture capital backed capital markets technology company building transformative prediction systems for fixed-income market participants. He is an Editor of Management Science in the Finance Area and an Associate Editor for Operations Research, Mathematical Finance, Finance and Stochastics, and other journals. Kay has won the JP Morgan AI Faculty Research Award (2019), the Fama/DFA Prize for the Best Asset Pricing Paper in the Journal of Financial Economics (2011), and the Gauss Prize of the Society for Actuarial and Financial Mathematics of Germany (2003).