Ashwin Rao, Vice President of Data Science & Optimization at Target
Pricing and Hedging of Agency Mortgage-backed Securities (MBS)
We begin with a brief overview of the agency mortgage market, covering the participants (the agencies, investors and borrowers), securitization, structured products, and an emphasis on the key risk of borrower prepayments due to refinancings and relocations. We will give an overview of the industry standard (on most trading desks) for pricing/hedging agency MBS: a stochastic interest rate model (calibrated to interest rate derivatives), a regression model for mortgage rates, a hand-crafted econometric model of prepayments, a cashflow-generation library, Monte-Carlo simulations, and pricing/greeks based on the much-misunderstood notion of “Option-Adjusted Spread” (OAS). In an attempt to overcome the pricing/hedging challenges experienced by traders operating in the above-mentioned framework, we will highlight alternative formulations with mathematical/computational appeal: (A) Martingale-based MBS Pricing formula and the equivalent PDE, that help with intuitive reasoning regarding valuation/risk (B) Closed-form approximations as well as Backward-Induction Pricing, (C) Demystification of OAS by expressing it as the Price of Model Risk (POMR), and (D) A prepayment model calibrated to MBS traded prices that organically captures POMR and hence, produces prices and greeks that are not just technically appealing but also more appropriate for usage on trading desks.
Ashwin Rao spent most of his 14 years on Wall Street as a Trading Strategist and Quant Modeler on the Interest-Rates and Mortgage Trading Desks of Goldman Sachs (as Vice-President) and Morgan Stanley (as Managing Director). In 2012, he founded a tech startup (named ZLemma) that developed products to match job-seekers and employers. ZLemma was acquired in 2015 by Hired, Inc. to whom he is currently an Advisor. He is currently also the VP of Data Science & Optimization at Target Corp. where he focuses on Stochastic Optimization, Large-Scale Simulations and Reinforcement Learning, mainly applied towards Stochastic Inventory Control for Target’s next-generation Supply-Chain and Logistics. Separate from his professional commitments, Ashwin writes code for Algorithmic Trading on an ongoing basis, and is currently experimenting with recently-developed Machine Learning techniques. Over the past decade, Ashwin has taught about 25 short courses (of 1-10 hours each) at universities and corporations with topics ranging across Finance, Pure & Applied Mathematics, and Computer Science. Ashwin has a B.Tech in Computer Science from IIT-Bombay and a Ph.D. in Algorithmic Algebra from University of Southern California.