Analysis of Mortgage Backed Securities: Before and after the Credit Crisis

Abstract

Valuation of mortgage backed securities (MBSs) and collateralized mortgage obligations (CMOs) is the big science of the financial world. There are many moving parts, each one drawing on expertise in a different field. Prepayment modeling draws on statistical modeling of economic behavior. Data selection draws on risk analysis. Interest rate modeling draws on classic arbitrage pricing theory applied to the fixed income market. Index projection draws on statistical analysis. Making the Monte Carlo analysis tractable requires working with numerical methods and investigation of a variety of variance reduction techniques. Tractability also requires parallelization, which draws on computer science in building computation clusters, applying new technology such as graphical processing units (GPUs), and analysis and optimization of parallel algorithms.

Here we detail the different components, describing the approach we have taken in each area. Of particular interest is how the credit crisis that started in 2007 has impacted the modeling.

The end result is that accurate price calculations on individual securities can be done in real time, and the entire universe of CMOs and MBSs can be analyzed overnight.

Keywords: MBS, CMO, OAS, credit crisis, subprime crisis, interest rate modeling, rate, yield, yield curve, Gaussian, Monte Carlo, parallelization, GPU, CUDA, Markovian, mortgage, mortgage backed, collateralized mortgage obligation, collateralized, structured product, prepayment, prepayment modeling

JEL Classification: G12, G13, C15, C51, C52, C61, C63