Yet another chapter, which pretty much gives an example of using the operators described in Chapter 12 – Three Tools.
The actual path to a finished structure has an extra fifty or more iterations as each potential major bond buyer asks for “tweaking” and we adjust face amounts, schedules, etc. Also, when a deal is in the final stages of structuring, it can change because every twist and turn of the yield curve changes the economics.
Remember, we did these deals for a total profit equivalent to only two or three basis points of yield. When markets sometimes move twice that much in a day, the fine-tuning can be quite maddening.
You should be able to see how a structured deal with multiple kinds of investors works by walking through this example.
Anatomy of a Deal
In the bond world, there are a few customers that have to feel like they got a better deal than anyone else, and they often ask to see the one bond class (not theirs) that’s terrible.
Other customers really want to understand everything, so giving them lots of information helps them make their buying decisions.
When I worked as a deal structurer, I tended to avoid the first kind of customer and really enjoyed working with the second kind.
Some investors need to see the downside, and figure out how a bond they’re considering can survive the stress. A few need to see that they will have big upside in a scenario that could reasonably occur, like people who bet on long shots at the racetrack.
Long-term successful investors tend to start with the structure of their liabilities, and look to capture a profitable spread over the cost of those liabilities under almost every possible future performance “path” for the asset (in this case, a bond) being acquired.
The other approaches, like avoiding risk at all cost, or looking for big upside, tend to be less successful over the long run, simply because you can’t always predict either the best or the worst of potential future scenarios, except to say that neither is very likely.
In the late 1980’s, most of our deals were done with Agency (Ginnie Mae, Fannie Mae, or Freddie Mac) guaranteed mortgage pass-throughs as the underlying assets that provided the raw material for the cash-flow engineering that was the domain of the deal structurer.
In structuring CMO’s, the main risks that were divided up were interest rate risk and prepayment (maturity) risk. When we created bonds with principal payment schedules that had preference, we divided the prepayment risk among the various bond holders.
Structuring to address the risk of prepayments became even more important to investors after the surge in prepayments in 1986, when Ginnie Mae MBS got prepaid so quickly that they stopped being thought of as ten-year bonds because they behaved more like five-year bonds.
When rates went back up towards the end of that decade, mortgage investors may have felt like they had whiplash as prepayments slowed down again. By 1989, the ten-year bonds that acted like five-year bonds looked like they were going to act like ten-year bonds again.
It’s important to remember that when structuring a deal, you have to set up all the payment rules today, and those rules can’t be changed later as conditions change. Some investors want “short” bonds only, so they are sensitive to extension risk. Others want “long” average life bonds. Still others want floating rate bonds that they can finance at a profit whether rates move up or down.
After the prepayment whiplash between 1986 and 1989, some investors steered clear of mortgage bonds completely until they were assured that their investments would not vary even if prepayments came in twice as fast or half as fast as originally expected.
Extremely risk-averse investors decided they would only invest in mortgage bonds that paid on a fixed schedule. Part of our structuring job was to create CMO bonds that would give investors a way to avoid wide variance in the time it took to get their principal payments.
The deal structure described below and steps in creating it are an abbreviated version of the actual process we follow when we create CMO deals with customized bonds. The deal most similar to this one was JHM Acceptance Series E, launched in March of 1989. The collateral was GNMA 9’s.
When I worked as a deal structurer, I also used diagrams to show salesman, traders and investors what the deal structure looked like. The diagrams in this chapter are the same kind of diagrams I used on Wall Street, and are stylized versions of what it looks like when you “map out” the cash flows in a securitization.
Diagram A – Mortgage principal payments at expected prepay speed
They also demonstrate, graphically, how the desires of many different investors are met in a single deal. Some investors could take the risk of maturity variance, while others insisted on relative certainty. One investor was looking for something special – a bond that didn’t shorten in fast prepay environments, but actually lengthened, even though it did not adhere to a strict payment schedule.
I always worked closely with investors to meet their needs as I structured the deals. I was pleased when I could create great bonds for every investor, by making every tranche an ideal fit for the individual investor’s needs. There was a great feeling of satisfaction when it all came together, the same kind of satisfaction a mathematician has when he writes an elegant proof.
First, I’d map out the expected principal payments, and then I’d run out the cash flows at fast and slow prepay speeds. That would give me visual patterns of principal payments under a number of different scenarios. The “rules” for how the cash flows are allocated are embedded in these diagrams, which show a “snapshot” of a deal and the range of behaviors of all the bonds in it, based on those different scenarios.
Diagram B – Principal payments at speeds twice as fast and half as fast as expected
Having mapped out the reasonable fast and slow principal schedules, I set up the tranches called Planned Amortization Classes (PAC’s) to satisfy the demand for stable bonds that are protected from prepayment variance.
As Diagram C shows, the PAC bonds are designed by setting their schedules as the minimum principal payment amount for both the fast speed and the slow speed. Because the PAC bonds have priority to hold their schedules, investors in the PAC bonds give up some yield.
Diagram C – Defining payment schedules for the PAC bonds using fast and slow speeds
The next diagram shows the “first cut” of the bond deal as it evolves. It has a short high yield bond (1.1 to 1.2 year average life) that gets some of the excess yield the PAC bonds give up, but absorbs the prepayment variance. If prepayments come in slow, that one-year bond might extend to being a three-year or four-year bond. The big middle part of the deal is the “Total Return” or TR bond, which will eventually have the unusual property that it extends in duration when the mortgages prepay faster.
Diagram D – Laying out potential “companion” bonds to absorb the major variance in principal payments, we have a very short bond (Class A) that will be bought by an insurance company with short liabilities that wants the extra thirty basis points of yield, and will take the extension risk
Now it’s time to test to see if the TR bond and the long floater/inverse pair perform acceptably in a fast prepayment scenario. The floater and inverse floater together form a single fixed-rate bond, so they will always pay in parallel, in the ratio they began with. The new TR bond does expand to take the long cash flows, but in this deal structure, too much early principal payment comes to the TR bond, so I’m not getting the average life lengthening that I’d hoped for.
Diagram E – In the fast prepay test, we see that the TR bond begins getting paid in year 3, so even though it gets the later payments after all the other bonds have paid off, it does not extend its average life enough to get the customer the extension they want. The PAC protection needs to be less strong, beginning by removing the P4 bond, and changing the rules for P3.
I’m not too concerned about the floater getting longer or shorter, because the investor who wants it is a bank that will be issuing CD’s every month to fund the bond according to how much principal is still outstanding. Whether it pays off quickly or slowly makes little difference to this investor, as long as LIBOR stays below the 11.5% effective cap.
The inverse gets a yield of 15% to 16% if things don’t change, and because it sells for a discount, it yields even more if it pays off quickly. Its main risk is if LIBOR goes above 9%, because that will cause the bond to have a yield of only 7.5% to 8%. If LIBOR goes to 10%, then yield will drop several hundred basis points more.
Now it’s time to fine-tune the structure. I’ve changed the PAC series to eliminate the P4 class and allow the five-year P3 class to take some of the prepayment risk that the TR buyer is paying to avoid. The back-end floater-inverse pair of bonds and the Class A bond still take most of the prepay risk. The short Class A bond is larger to help with the fast prepayment absorption.
Looking at the schematic below (Diagram F) at the “pricing speed” (expected prepay speed), we see that the TR bond takes some of the early cash flows and stops paying around year twelve if the mortgages pay off as expected.
Diagram F – The deal at the prepayment speed we expect the collateral to pay, also known as “pricing speed.” The PAC series has been modified by removing the longest PAC (P4) and by changing the allocation of prepayments. Class A and the TR bond have both gotten larger.
The TR bond is now a bit larger (it’s nearly 25% of the deal) and I’ve changed the payment rules to allow the TR bond to be the last bond paid if an “overflow” of principal needs to be allocated. With these changes, the Class A bond has grown to about 1.4 years’ average life from the shorter amount in the earlier versions of the deal, and the TR bond and floater/inverse pair have each gotten a little shorter.
Diagram G shows how the bonds pay in a fast prepay environment. The Class A and the floater/inverse pair absorb a lot of the extra early principal, but the P3 bond also takes a small amount of early payment. The TR bond now extends to collect the entire last twenty-five years’ of the mortgage cash flows, giving it an extension of about two years even though the underlying mortgages are paying off twice as fast as expected.
Diagram G – Here’s the deal with prepayment speeds twice as fast as expected. Class A has shortened to 0.75 years from 1.4, the floater/inverse pair is just 2.2 years long, and the P3 bond has lost less than a tenth of a year in average life. The TR bond is now nearly 3 years longer than it was at pricing speed.
In designing the TR bond in the deal that’s simulated here, I was working with a customer who worked so closely with me in the process that he actually served as co-designer. That investor bought the bond he co-designed, and paid approximately one percent more than he would have paid for a bond without its unique features.
The last set of changes in the deal structure results in a slight risk of shortening for the P3 PAC bond, but the amount it shortens under fast prepayment is so slight that the buyer will only need a couple of basis points of extra yield. We can provide that by lowering the price of the bond just one eighth of one percent.
The amount of average life extension the TR bond enjoys in the fast prepay scenario is significant. It now gets several years longer when the principal comes in fast, and the customer for this bond is willing to give up yield to enjoy that “option” behavior that counteracts the normal behavior of mortgage securities and callable corporate bonds. The yield the TR bond buyer gives up translates into higher yields available to reward the investors in the other three fourths of the deal.
I’ve just walked you through the process of structuring an MBS deal that’s far from plain vanilla, but not the most complicated of deals by far. This is an abbreviated version of the actual process that goes on. The real process takes days and dozens of iterations, with intervening conversations with potential buyers of bonds if they are affected by structural changes. Then there are the deal ideas that never go anywhere, or the deal structures that work one day, until the market moves, and they don’t work at all.
Back at Pru, we called our deal structuring room the “War Room” for a good reason, because the team in there was constantly dealing with changing market information along with a flow of requests and reactions to deal structures. We kept a big white board covered with requests for bond characteristics, and tried to fit the requests into each deal.
At the end of the day, the deal structure described above gave an attractive investment to each buyer in the deal. Furthermore, the deal was innovative, containing bonds with characteristics that were simply not available in the bond market at that time.
Agency Mortgage Securities – MBS or CMO’s that are guaranteed by Ginnie Mae, Fannie Mae or Freddie Mac. Timely payment of principal and interest are guaranteed. These organizations are also called GSE’s, for Government Sponsored Enterprises.
Near the end of the 80’s, after I set up the Mortgage Finance effort at UBS, we were able to adapt the structures of the Agency CMO business to non-guaranteed “private label” MBS, and for a while we had very little competition.
At that time, the insurance companies Travelers and Prudential were major issuers of non-Agency MBS, as was the General Motors finance subsidiary, GMAC. We used the AAA senior bonds from these issuers as our raw material for new CMO deals, creating the same kind of complex CMO structures others were creating from Agency MBS.
The mortgage loans behind these AAA bonds were also called “jumbo” loans. They were as good or better credit quality than the loans in the Fannie Mae or Freddie Mac deals, but they didn’t qualify for Fannie or Freddie because they were above the upper size limits for Agency guarantees.
Since the cash flows from those “private label” MBS bonds had very minor differences from those used in the loans backing Agency MBS, a deal that used AAA bonds as collateral had to capture those subtle cash flow differences in order to comply exactly with the legal disclosure documents.
Whole Loan Mortgage Securities – MBS or CMO’s that have mortgage loans rather than guaranteed mortgage pass-throughs as their collateral. Without the guarantee, the only source of payments is the loans themselves, so rules are set up to allocate delinquencies and defaults by the borrowers rather than assuming all payments are on time as they are when a guarantee is in place.
While the competition played catch-up, we had about 40 to 80 basis points of excess yield to work with to create customized investments in full-featured CMO structures built from “whole loan” MBS. By the simple act of combining the more complex structures from the Agency CMO business with the extra yield in the AAA piece of the whole loan deals, we could give our clients the same kind of tailored cash flow bonds they bought in Agency CMO’s, except with higher yields. We could also pay the sellers of those private label MBS about a quarter of a percent more for their AAA MBS than the rest of the market.
We did it all using a single computer model as our tool. By another of the industry’s historical accidents, most Wall Street mortgage departments had two sets of computer models (along with two sets of traders), one to deal with Agency mortgage bonds, and the other to deal with private label (whole loan) deals.
It took almost a year for most of the competitors to combine their two models to catch up to us. That year was 1989 and after only months in the business, our innovations led us to attain the number one position in the “private label” CMO bond business.