Subprime Mortgage Credit Derivatives

CHAPTER 10

Loss Projection for Subprime, Alt-A, and Second Lien Mortgages

In this chapter, we provide a detailed, completely transparent approach in projecting collateral cumulative losses. We outline and then expand on a simple loss projection model that is based on what we call default pipeline and default timing curve. We also provide an alternative specification of the simple model, which further allows users to take the effect of prepayment into account in loss projections. Finally, we elaborate on the slight differences between applying our methodology to subprime, Alt-A, and second lien collateral.

TWO WAYS OF PROJECTING LOSS

There are two ways of predicting collateral losses. The first way is through a full-blown econometric/statistical model. Typical inputs are collateral characteristics, namely, FICO, LTV, documentation type, loan purpose, loan size, and so on. Macroeconomics factors, the most common being home price appreciation (HPA) and interest rates, are also model inputs and need to be forecasted. There will also be interactions among various factors. For example, HPA may have a larger effect on high LTV loans than on low LTV loans. Some effects are age-dependent (e.g., FICO effect on default seems to dissipate over time). Many effects could also be nonlinear. Some models also include performance history (e.g., delinquency status) to forecast future performance.

An econometric model is estimated by running many regressions on historical data. There will be out-of-sample ...