As capital market consultants, we spend a lot of time talking and thinking about servicing-what it is, how it is valued, and how it can be expected to behave.  While a lot of people in the industry have an understanding of servicing and its related securities, I think it will be helpful to take a basic look at the concepts underlying assets comprised entirely of interest cash flows, as well as how that market evolved.

Let's first take a look at mortgage cash flow streams. As we know, mortgage cash flows are comprised of principal and interest, both of which behave quite differently.  The key point is that the amount of principal to be returned from an agency pool is fixed; prepayment speeds only influence when the cash is returned.   However, timing does impact the value of the principal cash flows due to the time value of money.  As a simple example, the principal component can be envisioned as a series of zero-coupon cash flows that an investor is scheduled to receive over time.  If the timetable for the payment series is accelerated, the investor makes a windfall profit; alternatively, slowdown in the receipt of the cash hurts the investor's returns.  

The interest component is entirely different.  Unlike the principal cash flows, slower prepayment speeds benefit the returns on a pure interest cash flow, while faster speeds hurt its returns.  This is because the amount of interest to be paid to investors is a function of the amount of principal outstanding, which in turn is dictated by prepayment speeds.  (Think of the interest as a dividend thrown off by the business; if the business shrinks or disappears, so does the dividend.)

To illustrate this difference, let's take a brand-new FN 4% pool with an original face value of $1mm.  Changing the prepayment speed impacts the timing of the principal cash flows, which is reflected in the security's weighted average life or WAL.  (The WAL is the weighted amount of time that the bond's principal is outstanding, given some prepayment assumption.)  If a 0% PSA prepayment assumption is used (i.e., there will be no prepayments on the pool), the WAL is 18.3 years; the WAL declines to 8.95 using a 150% PSA assumption.  However, in both cases the investor receives a total of $1mm in principal back.  (Remember that this is an agency pool, so our favorite Uncle guarantees the principal.)  However, the total amount of interest paid varies greatly with the prepayment assumption.  At 0% PSA, the investor would receive $732,500 in interest; at 150% PSA the total interest received drops to $356,486.  Therefore, prepayments impact both the timing and the amount of the interest cash flows paid to investors.

Investors first tried to capitalize on the different behavior of mortgage principal and interest by creating securities that redistribute interest within the security (or, more accurately, a trust backed by MBS pools).  The earliest "strips" were created by taking current-coupon pools (then 9%s) and then creating "discount" and "premium" securities with coupons of, say, 6% and 12%.  These eventually evolved into so-called principal-only (or PO) and interest-only (or IO) securities.  Markets for IOs and POs developed for a variety of reasons.  Some investors that were looking to make explicit bets on the direction of prepayment speeds found either one or the other side to be attractive.  In addition, some investors took advantage of the leveraged performance of the bonds to make explicit interest rate bets; finally, both POs and IOs can serve as hedging vehicles.  (Also, investors looking to create synthetic discount or premium securities can re-combine IOs and POs from the same trust to create a variety of coupons.)

Given the sensitivity of IOs and POs to varying prepayment rates and the fact that prepayment speeds are primarily influenced by interest and mortgage rates, the price behavior of IOs and POs are quite different.  POs tend to have durations that are greater than the underlying pools (or "collateral," using the industry terminology); the normal response of a bond to changing rate levels is amplified by the associated change in speeds.  By contrast, IOs typically have negative durations, i.e., their prices rise as interest rates rise.  What's really happening is that the normal price sensitivity of any cash flow based on changing interest rates or IRRs is offset, to varying degrees, by the fact that the composition and value of the cash flow is strongly influenced by the underlying asset's prepayment speed.  (Note that IOs don't always have negative durations.  IOs backed by loans that are very far out-of-the-money, for example, may have a flat or slightly positive duration; since changes in the prepayment speed of the collateral will probably be muted, the basic sensitivity of any cash flow to interest rate changes will predominate.)

This leads to a discussion of the relationship between servicing assets and IOs.  Excess servicing (i.e, the interest cash flows that falls out as part of the process of pooling) is, truly, an IO cash flow; excess servicing can (and has) been securitized into so-called excess-servicing IOs.  Required (or "base") servicing, however, is significantly more complicated to value and hedge.  For example, it costs more to service delinquent loans; as with non-agency securities, some estimation of future delinquencies and defaults must be made in order to accurately gauge the potential future costs of servicing a loan.