11. Additional questions related to the cash flow model from question 8:

a) Howhigh does the collateral default rate have to go before the ClassBsuffers a reduction of yield? Is it the same in the pro rata and sequential cases? At the collateral default rate where the Class B loses yield, how much yield is lost?

b) The reduction-of-yield method relates the average loss of yield on a security, produced in a Monte Carlo (MC) simulation, to the alphanumeric rating. If the yield reduction obtained at the default rate in question 11a were the result of an MC simulation (average reduction of yield) you could obtain a rating from that number. Find the rating by looking up the reduction of yield (in basis points) in table 9.2. Using the credit spreads in table 7.2 below, adjust the Class B interest rate input to reflect this rating. Is the impact on the Class B rating stable or does it change the rating? What is the significance of question 11b?

c) Given the parameters in question 8 and the outcomes, what is the seller’s

“optimal” Class B size? Why can’t you make this the real Class B size? In general, given a rating on a security, how do you find the buyer’s optimal size?

d) How would you adjust the original scenario to analyze a securitization of subprime autos with a five-year WAM?

e) It turns out that, underlying the collateral are two distinct borrower populations, each with its own characteristic loss curve. How would you factor this intelligence into your analysis?

f) In the scenario where defaults are 22%, the originator/seller/servicer eventually defaults in time step 55 and a backup servicer must step in. Is this a good deal or a bad deal for the backup servicer? Given your answer, can you generalize about pricing for backup servicing?

g) Rating agencies do not model the base case loss curve in cash flow models.

In the 1990s, they stressed the loss curve by ramping up losses from 0% to fully loaded over a 6-, 9-, 12-, or 18-month horizon. How much capital did their stress cost the structure in question 8? Is stressing the input assumptions a good idea?