time value of all the embedded options (including caps) and, hence, the overall option cost. As for the loan features (loan size, LTV), we find them to play a rather marginal role in jumbo values.

The Sub-Primes Fare Well
The new non-agency sub-prime ARM model relies heavily on the SATO factor. Prepayments are still interest rate dependent (albeit, to a smaller extent than the prime ARMs), but this dependence comes with a smirk: lowering market rates by 100 bps have an incomparably stronger prepayment effect than raising the coupon (and GWAC) by the same amount. Since SATO is employed both in the refinancing incentive and credit impairment, the new prepay pattern becomes less dependent on the original coupon.

On a pure prepayment basis (without consideration of losses and difference in margins, i.e. zero-SATO), sub-prime ARMs are better-behaving, value-adding instruments than agencies. For a hypothetical 3/1 sub-prime 6.5% hybrid (usually called “3/27” -- most sub-prime hybrids have 2-yr or 3-yr intro periods), this favorable pattern translates into a 1.0 point pay-up to the agency (0.5 point for 5.5%). Each $100K increment in loan size reduces the value by 0.08-0.12 points. The effective duration’s extension is sizable: 0.8 yr for 6.5 and 0.5 yr for 5.5. With the original GWAC rising, refinancing ability of sub-prime borrowers becomes saturated and effective convexity becomes positive.

On a full-value basis, an approximate relative value exercise can be carried as follows. Fix an agency coupon rate (let it be 5.5, GWAC = 6.0, LIBOR + 175) and compare it to a family of sub-prime ARMs with various SATO levels. For example, SATO = 1 means a 6.5 coupon, 7.0 GWAC and a 275 net margin. Furthermore, we should add 100 bps to OAS to account for the credit difference. Results are below.

Table 1. Relative valuation of sub-prime 3/1 (3/27) hybrids