I still have some irons in the fire for the New Year that may cut down or cut off my blog posting, but I did promise to share a favorite chapter from my Mortgage Market Mayhem manuscript.
I will actually share two of them, once I figure out how to import hand-drawn diagrams into this blog editor. Both chapters deal with the process of structuring deals, one of the most reviled, and least understood, activities in modern finance.
When I first started doing this in the mid-80’s, it took a while before I could figure out how to tell my friends what I was doing. They automatically assumed that I could give them stock tips because I was working on Wall Street.
Nothing could be farther from the truth. During most of my years on the Street, I was too busy to even know (or care) where the Dow was trading. My own money was in money market funds so I never had to deal with compliance issues.
Having been an avid stock and option trader before I came to the Street, it was quite a change for me.
Anyway, the description I hit upon (ca 1985) to describe what I was doing was “financial engineering.” As far as I know, I was the first person to describe my profession with that phrase. Ten years later, there were business schools offering that as a major, but I don’t think they had any deal structurers teaching the courses.
This chapter, and the next, will show how I did that job.
Three Tools is All It Takes
This chapter presents an overview of the three basic methods of tranching for asset-backed securitizations, and provides a conceptual foundation for understanding how securitization allocates risk. For the individual or investment professional wishing to explore each method in depth, the overview is followed by an in-depth examination of each tranching technique. The text quotes extensively from a chapter I wrote for the 2006 Euromoney International Debt Capital Markets Handbook.
Securitization – the largest of all capital markets – caters to investors with vastly different risk appetites and investment needs.
When a deal structurer sets out to design a securitization to satisfy investor demand, that structurer needs only three basic tools, or operators, to create those new securities. Those ways of dividing, or tranching, the cash flows from the underlying assets are:
- Time tranching
- Credit tranching
- Coupon tranching
Three operators are plenty. After all, every mathematical expression only needs the operators of addition, subtraction, multiplication and division, and those four are enough to make an infinite number of equations. (Before the purists get upset, I freely admit that you can define multiplication as the repeated application of addition, and you can define subtraction as the inverse of addition.)
Any securitization is fundamentally a form of redistributing risk; it does not lower the total risk embedded in the loans or leases that form the “raw material” for the securitization process. However, the portfolio effect of having thousands of obligors and the liquidity introduced by capital markets trading does lower shorter-term risk to an investor who might otherwise be investing in a small number of unsecuritized credits. After looking at the whole picture, securitization lowers the value at risk (VAR) for all but the largest investors for these assets.
Even the largest investors may find the liquidity and standardization of securitized bonds advantageous when compared to the cost of maintaining a staff and taking on the expense of purchasing loans separately in an unsecuritized form.
All multi-tranche securitizations involve setting up payment priorities among the classes, and a pre-defined set of rules for those payments. Since the rules cannot be changed later, the goal of the investor and the structurer is to weigh the potential outcomes while taking into account all the future events that might happen. The most successful investors do not allow a single bad scenario to drive their investment decisions, nor do they let themselves be drawn into an inappropriate risk posture just because they see a potentially very positive outcome.
“Time tranching” divides cash flows into individual tranches (bonds) which may be paid sequentially, in parallel, or by defining variable proportion cash-flow allocation based on any number of factors. Such factors could be simple, such as the time since the deal closed, or more complex, through an algorithm driven by a calculated variable such as total pool balance, pool performance, or even some external independent variable.
Another version of the time-tranching technique involves setting a principal payment schedule for some bonds and letting other bonds, commonly known as “support bonds,” take the variance in principal payments.
The very first CMO, issued by Freddie Mac in 1983, was a sequentially paid, time-tranched bond. In that deal, known as Freddie Mac Series A, there were four bonds. Only when the mortgage pool had paid enough principal to retire the first bond, did any principal payments go to the second bond. Then the second class received all of the principal payments until fully paid, and so on through the fourth and final class.
Since all the principal for the underlying mortgages was already guaranteed by Freddie Mac in the form of pass-throughs, there was really no change in the aggregate credit exposure for investors, except for the fact that the first class to be paid had a shorter exposure to Freddie Mac’s credit guarantee. In the jargon of the business, this deal structure came to be called “plain vanilla sequential,” or “ABCD sequential.”
Later time tranching (or principal allocation) developments included setting principal schedules for some bonds and giving them a schedule of payment preference (rather than simply a payment preference). By having a principal payment schedule, one group of bonds in the securitization structure was less likely to be paid off before the investor expected to be paid off, even if the underlying mortgages paid off faster than expected. Though generally not as large a concern, some market conditions, like interest rates going up, also made investors worry about not getting paid as soon as they originally expected. This “extension risk” is just the other side of the coin of early payment. The unscheduled “companion” or “support” bonds also absorbed some of the extension risk along with the early payment risk.
Support Bond – A bond that absorbs undesirable characteristics like credit losses or prepayment variance so other bonds are more attractive investments.
The scheduled bonds, called Planned Amortization Classes (PAC’s) or Targeted Amortization Classes (TAC’s) gave their investors less exposure to the borrowers’ prepayment option. Investors in PAC’s and TAC’s give up some yield in exchange for this lower maturity variability, and that extra yield is available to reward the investors in the classes that take on the extra risk.
Another common method of “time tranching” is the technique of allocating principal payments to different sets of bonds at different ratios as time passes. One popular bond structure started out using this variable principal allocation method as a way of protecting or “supporting” senior credit tranches. The senior bonds got a disproportionately high percentage of the principal payments for the first few years while the junior bonds weren’t paid, so that the credit protection for the senior bonds grew over time. The technique was called “accelerating” the principal payments, especially when the senior bonds got 100% of the principal for the first three years.
When this time tranching technique migrated to the fully guaranteed Agency CMO market, the bonds came to be known as Accelerated Senior (or “AS”) bonds. The other bonds, called “NAS,” for “Non-Accelerated Senior” bonds, were protected from fast prepayment risk the first few years. The NAS bond takes a disproportionately low percentage of prepayments early in its life, and after a certain date, gets a disproportionately large percentage of the principal.
A different method of dividing the cash flows is through credit tranching. These bond structures are paid according to their credit priority, often using “time tranching” as described above on the most senior AAA rated bonds.
The typical way to allocate credit losses in these deals is to allocate any losses first to the reserve funds, excess interest cash flows, insurance policy or overcollateralization that might be included in the structure to provide support. If any such reserves, excess cash, insurance or extra collateral have been exhausted by covering losses already, then the credit loss is allocated to the lower priority tranches as principal by writing off some of the balance of the bonds. In other words, the investors in the junior bonds lose some of their principal when a loan defaults, while the money salvaged by selling the bad loan or foreclosed asset is allocated to senior credit tranches as early payment of principal.
Because of these dynamic credit protections, many bond deals can be more robust than they appear at first glance. For example, a bond that has only 4% subordination to protect it (sometimes called “hard” protection) might appear on the surface to have very little protection against credit losses.
However, if that deal also has 300 basis points of excess interest available to absorb credit losses, then 5% in losses in a year would only use up 2% of the subordination in the first year, because the other 3% in losses would be absorbed by the excess interest. If each foreclosed loan resulted in a loss of half the loan balance, the 5% in credit losses would come from having 10% of the mortgages going into foreclosure.
From this, we can see that despite the seemingly modest 4% credit protection for that bond, 20% of the loans in the pool would need to default in the first two years before any principal would be lost by the bondholders.
For a typical subprime mortgage deal in 2005 and 2006, the subordination level required to achieve AAA ratings was around 20%, meaning that 80% of the deal would be rated AAA, while the remaining 20% would have lower credit ratings. Using the same basic assumptions from the example above with 3% per year in excess interest available, roughly half the borrowers would have to default and losses on each default would have to run 40% or higher before the AAA bonds would lose any principal.
Among the bonds in credit layers beneath the AAA bonds, the credit markets have developed demand for every gradation of credit tranche, so the resulting lower-rated subdivisions of the bond structure are so “thin” that each specific bond will be very sensitive to small changes in credit performance once the initial buffers of excess interest and subordination have been used up.
Decades ago, the credit ratings were single letters (A, B, C, etc.), but now each letter grade is divided into three “steps.” If the single A credit rating for a pool of mortgages requires a subordination of 7.5% while a double A rating requires 9%, then each step in the single A credit would only be 0.5% of the whole deal. Bearing in mind that the mortgages are paying 3% per year in excess interest, it might take actual losses of 13% to 15% over several years before the first dollar of loss is assigned to the A- (A minus) rated bond. Once that occurs, it only takes another half a percent in losses to wipe out that bond and begin to take write-offs on the A rated bond. Another 0.5% in losses would begin to eat away at the A+ rated bond.
If you were to analyze the collateral mortgages at the time a deal is put together, it might be difficult to determine whether the default rate in the third through the tenth year would be 8% per year or 8.25% per year, but that relatively small difference might make the very large difference between one of those A- rated “mezzanine” bonds paying all of its principal, or none of it.
Dating back to the days when bonds were elegantly engraved and printed on paper, a bond that had regular payments of interest had small detachable “coupons” printed at the edge. As each interest payment came due, the holder of the bond would physically present the payment coupon for that interest payment to the bank that acted as Payment Agent for the bond issuer. Because of that tradition, we came to call all interest payments “coupons,” even though the bonds themselves and their payments are now electronic entries.
Coupon tranching, the third method of dividing cash flows, is based on interest payments. Some tranches can be set up to accrue current interest rather than paying it, diverting that accrued interest to other classes as principal. The first of these interest payment tranching structures in the MBS market was the Z-bond, or accrual tranche, introduced in Freddie Mac’s Series C deal, launched in January of 1984.
The fourth class in that deal was a 12.1% bond that accrued its interest (added to principal) each semi-annual bond payment period, making additional principal available to pay off the three sequential classes which were time-tranched to be paid off before the Z tranche.
The Freddie Mac Series C Class Z bond compounded, or accrued, at the rate of 6.05% every six months, growing from an original balance of $80.2 million in 1984 to a peak of $211 million by July of 1992.
All of the interest that was added to the Z-bond balance was paid as principal to earlier classes in the sequential time tranching of the deal. The holders of those shorter maturity bonds had an extra $130 million in principal payments that did not rely on the underlying mortgage pools having early principal payments, so the shorter bonds in Freddie Mac CMO Series C got a much less variable principal payment than they would have otherwise.
Investors in the early classes in this $525 million deal had significantly lower exposure to the borrower prepayment option because of this innovation in interest payment tranching.
Some tranches are interest-only (IO) and principal-only (PO) bonds. The technique of dividing mortgage interest payments into IO’s and PO’s (interest-only and principal-only securities) became infamous in 1987, when Merrill Lynch’s Howie Rubin took the PO side of the trade and Franklin Savings’ Ernie Fleisher took the IO side of the trade in the billion-dollar deal known as Merrill Lynch Trust 13. Unfortunately, Howie paid too much, and Ernie made a killing.
Merrill took a very public $275 million loss on the deal. It wasn’t all Howie’s fault, though. Salomon Brothers, Howie’s prior employer, decided to price a similar IO-PO deal, selling their PO bonds for nearly 20% less than Merrill was asking.
In Wall Street’s revisionist history, Howie was blamed for making an “unauthorized trade,” but I have to say that I never saw a trader responsible for getting a prospectus printed for a CMO deal. It was a deal, not a trade, and it required the involvement (and signatures) of a number of bankers and analysts, along with senior management.
Post-fiasco, Rubin got a nice “kiss goodbye” from Merrill, followed by a decade of huge paychecks at Bear Stearns, which was only too happy to have the young trader who featured prominently in the book Liar’s Poker.
At about the same time, Goldman was also doing a variation of these IO-PO deals, much less publicly. Goldman’s preferred trade of the time was to take GNMA 9’s and create equal and parallel bonds, one tranche with a 12% coupon and one with a 6% coupon, extracting a nice profit from two sets of buyers. At the time, GNMA 9’s cost around par, or 100% of their face amount. The 12% bonds that Goldman created sold for a higher price, and were called premiums, while the 6% bonds sold for a discount.
Par – The number of dollars of principal a bond will pay off when it matures. A bond sold “at par” is sold for 100% of this face amount.
Premium and Discount Bonds – Bonds that are priced above and below par (100%). Since all bonds pay the par amount at maturity, a discount bond yields more than its coupon as it moves toward maturity and the premium bond yields less than its coupon as it matures and the premium amortizes.
If you were bullish on bonds you would want to own the discount bond, or a PO. If you were bearish, your preference would be the premium bond or an IO. The bullish buyer would want a discount bond because they would enjoy a higher return if prepayments come in faster than expected, while the bearish buyer will choose the premium bond, believing that prepayments will come in slower. In general, prepayments come in faster if mortgage rates drop, and slower when they rise, so a structure like Goldman’s also allows investors to make a bet on prepayment rates.
Another form of coupon tranching allocates most of the interest from the mortgages to create a series of fixed rate bonds that sell at par. The associated excess interest is allocated to another class with no principal. Those are called WAC IO’s, or excess interest bonds, because they don’t have a stated coupon, and only get the interest left over after their associated fixed rate par bonds have their interest payments satisfied.
Par Bonds – Bonds bought “at par,” so the yield approximately equals the coupon rate, no matter how long the principal repayment (maturity) may take.
Perhaps the most interesting and controversial coupon tranching technique has been to divide the payments of a fixed-rate CMO bond into a pair of bonds that have variable interest rates, with one group collecting floating rate interest payments referencing an external market rate like LIBOR, while another bond paired with it becomes an “inverse floater.” The floating rate bond and its companion, the inverse floating rate bond, were introduced in 1987.
Taking the example of 9% GNMA MBS again, we could divide the bonds into three parts LIBOR floating rate bonds (“floaters”) and one part inverse floaters to create two classes of bonds that pay in parallel to each other. If we let the inverse floater coupon go to zero if LIBOR goes high enough, then each floater can have a maximum coupon of 12%.
The LIBOR floater bond might appeal to a bank that funded itself at LIBOR, as long as the mortgage bond offered 50 or more basis points of interest above the bank’s funding rate. When we used GNMA collateral for the deals, the banks only had to reserve 1.6% capital against their holdings of these bonds because of the U.S. Government guarantee on GNMA’s, so a positive spread of 50 basis points on the funding produced a very high return on capital for investors.
The buyers of the other bond, the inverse floater, were getting a very high interest rate to start with. If LIBOR was at 6% and the LIBOR floaters were at 6.5%, the inverse floater was getting 9% (the rate on the GNMA’s) plus three times the 2.5% difference between the GNMA rate and the rate the banks were getting on their floaters, for a total of 16.5%.
With an initial interest rate that high and no credit risk on the GNMA’s, there will nearly always be a buyer willing to make the investment in an inverse floater despite the concentrated interest rate risk.
If LIBOR goes too high, the cost of borrowing at LIBOR could exceed the return from holding a bond that can only pay a maximum of 12%. That effectively kicks in when LIBOR hits 11.5%, since the spread in the bond will shrink from the original 50 basis points.
This has not happened in the twenty years these bonds have existed, but there have been times when LIBOR traded above 9%, and looked like it might go higher. When that happened, the holders of CMO bonds with effective caps at 10% to 11.5% suffered a loss in market value. If investors sold during those episodes, they lost money, just like holders of 6% Treasury Bonds lost money if they sold when Treasury rates were high. If the floater buyers held on, they continued to get their spread of 50 basis points over LIBOR and an attractive return on capital.
For the owners of the inverses, these episodes of high interest rates were truly dire. If LIBOR went to 10%, the floaters would get their 10.5%, but the underlying GNMA’s were still only paying 9%. The inverse owners got what was left over, which equaled 9% – 3 * 1.5%, or just 4.5%. Needless to say, getting only 4.5% when short term financing costs 10% or more can be painful, or even financially fatal.
The inverse floater buyers of the late 1980’s did well if they managed to hold on. Even though one-month LIBOR did go above 9% in 1989 and the coupon on those inverse floaters dipped below 7%, investors that kept their inverse floaters until the early 90’s ended up with coupons that went over 20%.
Around the same time that the floaters and inverse floaters were gaining in popularity, I got to introduce the super PO bond in a private-label, non-Agency CMO deal, American Southwest Series 63. It was designed to provide huge upside if prepayments came in fast, and slightly positive returns (around 5%) if prepayments came in slow. When the mortgages in the pool paid off quickly a couple of years later, the buyer of that first super PO got $10 million back on a $1.4 million investment.
Over the years, literally thousands of unique structured deals have been launched in the securitization market, with almost every imaginable underlying source of cash flows “feeding” the various bonds in the structures. To date, every bond type and bond structure has been made by some combination of these three structuring operators: time tranching, credit tranching and coupon tranching.
The success of securitization and structured finance comes from the power of these three structuring techniques to custom-tailor investments for a wide array of investors.
Banks may need safe, short maturity fixed rate bonds that match up with CD’s they’ve issued or they may need floating rate bonds to match up with their interest bearing checking accounts. Hedge funds need to offer the kind of high yields they can get from inverse floaters or from taking extra credit risk. Life insurance companies need bonds that won’t get paid off too soon.
Each can get what they need when the securitization market is working the way it should.