In the 4th Quarter 2003 release of SPV the four standard models have been re-estimated and we compare the estimation results to each other and evaluate the performance of the models relative to the performance of the equivalent models of the prior release.
In addition to the model update we examine the effects of introducing the price of a refinancing bond into the prepayment function.
Traditional required gain models assume that refinancing is done into a bond issued at par. As such they are unable to account for aversion mortgagors have to refinance into a loan with a low price, which would increase their nominal debt. Using data of new issuance for a broad range of bonds we fix, for each bond of the estimation sample a likely refinancing bond and use the price of that in the prepayment function. This is incorporated in such a way that the individual loan groups can react differently to the price of the refinancing bond.
Comparing the estimation results to a similar model not using the price of the refinancing bond we find that even though the new explanatory variable improves on the fit on a group level we find no significant series level benefits, especially not compared to what one can achieve by using a loan yield spread as is done in the government based models of SPV.