Valuation Commentary

AD&Co OAS 5.2b: Enhanced Data, Bloomberg Links, Pay-ups & More
by Alex Levin

Our July Pipeline introduced a new OAS version 5.2 loaded with improvements. Now, using this OAS as the engine, we have built a more powerful Excel application featuring the ability to take advantage of the enhanced pool data (loan size, LTV, FICO, geography, purpose, type, occupancy). The Excel sheet even has a Bloomberg link so that a user can download available data automatically. Whereas enhanced data is a benefit, it's not a must: missing data is disregarded and carries no effect on prepayments.

Simple Modeling Approach
How is the additional information being used? The entire set of data, beyond WAC and WAM, collapses into just two tunings: the turnover tuning and the refinancing tuning. For example, high average LTV leads to a slower refinancing, but a faster turnover. Loan size and credit score are other important indicators of refinancibility. Some states are proven to be faster than others. Occupants of single-family, first homes, refinance most often, etc.

Just one additional DLL comes with the system - it translates the enhanced data into turnover/refi tunings. All Bloomberg links are built-in, but always in a VBA module and never by worksheet formulas. Hence, one can overwrite or complement Bloomberg data without damaging the links embedded in the spreadsheet.

Exhibit 1. Enhanced information (grey cells) linked to Bloomberg


In Exhibit 1, we show the enhanced data loaded from Bloomberg for 4 Freddie pools. Computed tunings reflect available data; in particular, high credit score, low LTV and significant loan sizes somewhat inflate refinancing. Bloomberg is gradually expanding its downloadable information universe; currently only geographic data (4 leading states) comes back. We noticed that enhanced data can be downloaded for CMOs too, but only geographic information.

Valuation Consequence: Pay-up
Knowing the pool's specifics we can compute its fair value using either OAS or prOAS analytics. We can compare it to the matching-coupon, same-OAS (or prOAS), TBA price and determine a pay-up or, rarely, pay-down. We call the difference theoretical pay-up.

Pay-up can stem from low loan balances, high LTVs or state concentration. Contemporary MBS market offerings are full of new jargons: "LLB pay-up" (i.e. pay-up for low loan balances), "New York pay-up", etc. Pay-ups can also arise simply due to WAM and WAC difference from TBA assumptions which can lead to delivery arbitrages. For example, a broker would demand a slight pay-up (several ticks) for brand-new production of premium pools because their speed will likely be ramping up for several months. These market phenomena can be quantified using AD&Co OAS, version 5.2b.

Practical Pay-up
Market pay-ups are often smaller than theoretical ones; the reason being that many investors buy pools and place them into available-for-sale accounts. When selling the pool back to the market, they can't be assured of a fair pay-up. This market inefficiency leads to the following conservative assumption: a pool bought at a pay-up will be held for some short, investor-defined period and sold at the TBA price. What is the fair pay-up now given this adverse investment scenario? We define this measure as practical pay-up, which should be a function of the holding period, but is generally smaller than the theoretical pay-up.

How to compute the practical pay-up assuming we know the holding period, say 6 months? First, we compute theoretical pay-up assuming the regular, next-month settlement. Then, we compute theoretical pay-up using forward settlement in 7 months (our OAS system allows settling MBS as far as 12 months forward.) This is the pay-up we forfeit when selling the pool at the TBA forward price. By subtracting this forward pay-up scaled down for amortization and discounting from the next-month theoretical pay-up, we obtained the practical pay-up we seek.

Exhibit 2. Example of practical pay-up computation

Pool 
Description
Theoretical Pay-up
Expected Amortization
Discounting
 Practical Pay-up
Next month
7 months forward
Between months 1 & 7
6 months, 1 month forward
abc123
 New Pool
0.2
0.0
0.88
0.97
0.2
xyz123
LLB Pool
0.3
0.15
0.88
0.97
0.17

The last column in Exhibit 2 is computed as:

Practical pay-up = Theoretical pay-up (next month settlement) -
[Theoretical pay-up (7 months settlement)]*[Amortization factor]*[Discounting]

Pool abc123 draws its pay-up from being early on the age ramp; in 7 months there will be little difference between it and TBA. In contrast, pool xyz123 has low balances; its forward pay-up is still essential, though perhaps limited due to the steep forward curve making refinancing differences less important. In the end, a pool with seemingly higher theoretical pay-up may end up with a lower practical pay-up. Once again, we emphasize the knowledge of horizon as the key entry to this practical pay-up analysis.

The entire analysis, including processing massive dealer offerings, can be done fairly quickly using AD&Co OAS version 5.2b.