A Statistical Analysis of Delinquency and Prepayment Risk in Subprime Mortgage-Backed Securities

Abstract

The 2007-2008 financial crisis led to increased scrutiny over the ratings assigned by agencies to tranches of subprime MBSs and MBS CDOs. While the events of this recession demonstrated that these ratings were generally unreliable, analyzing the defensibility of these ratings is integral to the evaluation of the rating system. This paper examines the ratings assigned to these products by simulating the behavior of securitized subprime mortgages with a multinomial logistic regression model to pre- dict the monthly default and prepayment risks associated with individual mortgages and a fractional response regression model to predict the loss on loans given default. This analysis found that while AAA and BBB ratings can be assigned to tranches of subprime MBSs, the risk associated with BBB-rated tranches was underestimated. Additionally, we determined that the tranches of subprime MBS CDOs created from the BBB tranches of subprime MBSs should have never been rated AAA but could have received ratings better than BBB.