CHAPTER
TWENTY-SEVEN
THE EFFECT OF AGENCY CMO PAC BOND FEATURES ON PERFORMANCE

LINDA LOWELL

Principal
OffStreet Research LLC

Planned amortization class (PAC) bonds represent one of the largest sectors of the collateralized mortgage obligation (CMO) market. Among the factors expanding the market for these bonds were widespread defections from the corporate market by investors who were attracted by the high yields and triple-A credit quality and discouraged in their natural habitat by limited supply and tight spreads, as well as by event risk.

Typical PAC buyers tend to be asset/liability matchers such as life insurance companies and commercial banks, but the PAC market appeals as well to investors with active bond management strategies. These investors may have, for instance, opinions about the direction of mortgage-Treasury spreads or mortgage-corporate spreads, they may wish to execute barbell or other strategies designed to take advantage of expectations regarding the shape or direction of the yield-curve, or they may be seeking value advantages, for example between the PAC and other CMO sectors, or within the PAC sector, among PACs with different features.

The main attraction of PAC bonds lies in the fact that they provide a defined schedule of principal payments (or, similarly, target balances), which is guaranteed as long as prepayment rates remain within a specified range. (Hence the name, planned amortization.) Holders are insulated to a significant degree from the uncertainty regarding the cash flows of most MBS, which arises from the right of mortgage borrowers to prepay their loans at any time. The fact that the prepayment process is interest rate-sensitive—the tendency of homeowners to move or refinance is inversely related to the direction of interest rates—has a material impact on the average life, duration, and performance of mortgage securities. MBS shorten in rising markets and extend in declining markets. To compensate investors for taking this risk, pass-throughs and other MBS are priced at higher yields than other, noncallable bonds of similar credit quality (such as agency debt). PAC bonds partake of the incremental yield available in the MBS market while at the same time providing more certain cash flows. In addition to providing guaranteed payments within a range of prepayment behavior, PAC bonds can be further refined to concentrate payments over a shorter period of time or “window.” PACs with narrow windows are perceived to be better substitutes for corporate and Treasury bonds with bullet principal payments, and therefore typically trade at tighter spreads than PACs with wider windows.

This chapter is intended to serve both classes of investors—the buy-and-hold PAC buyer, who wants to partake of the attractive yields in the mortgage market but whose liabilities or actuarial requirements necessitate more stable cash flows than either pass-throughs or standard CMOs can provide, and the active portfolio manager. The chapter examines the different features of PAC bonds and their effect on market value and investment performance. Where it is possible to isolate a specific characteristic, an attempt is made to model and examine its impact on the average life and yield behavior of the bond, as well as on its theoretical or option-adjusted value.

THE TERM STRUCTURE OF CMO YIELDS

The yields that investors require for occupying different average life sectors of the PAC market are determined by the same factors that influence other fixed income investors: portfolio objectives and constraints, the current and anticipated shape of the yield-curve, expectations regarding the underlying monetary and economic determinants of interest rates, and so forth. In addition, PAC buyers require additional yield as they extend the maturity of their investments to compensate them for the increased risk that the prepayment collars may be broken, as well as the greater average life volatility of the later tranches in a transaction if the prepayment collars are broken. An indication of the average life volatility of PAC tranches of different nominal average lives is provided by Exhibit 27–1. The graph depicts the average lives of a series of PACs from a single CMO issue1 at two prepayment speeds extreme enough to break both the upper and lower collars of all the PAC bonds. The range between the average lives at the extremes gradually widens for longer expected average lives to its widest point among the intermediate-term PACs.

EXHIBIT 27–1
Extension and Shortening of PAC Bond Average Lives as Prepayments Vary between 0% and 600% PSA

image

As a result of investors’ demand for greater compensation for holding longer-term PACs, the generic CMO yield curves are more steeply sloped than the Treasury yield-curve. Generic yields for current-coupon PACs backed by current-coupon conventional collateral and for on-the-run Treasuries are depicted in Exhibit 27–2. Although a satisfying discussion of the issue is beyond the scope of this chapter, it should be noted that CMO spreads are also influenced by the same factors that affect pass-through spreads—volatility of market yields, prepayment expectations, supply of new product, and so forth. Normally CMO spreads track pass-through spreads, with the relationship enforced by the existence of CMO arbitrage opportunities. When pass-throughs cheapen relative to CMOs, new transactions are marketed, increasing the supply of CMOs and ultimately allowing CMOs to cheapen relative to pass-throughs, thereby reducing the arbitrage opportunity.

EXHIBIT 27–2
Illustrative Treasury and PAC Yield Curves

image

Other features of the structure such as the collateral coupon, the PAC’s coupon, the collars, and the window result in additional adjustments to the required yields. In addition, the cash-flow performance of the bond outside the collars is affected by other characteristics, such as whether its schedule is supported by accrual from a longer-term Z-bond positioned later in the structure, and its priority for receiving excess cash-flow. These characteristics determine how volatile the average life and returns of a PAC are outside the collars, and so also affect its marketability.

COLLARS AND COLLATERAL

The strength of the collars—whether they will be broken by actual prepayment experience—should be investors’ primary concern. The strength is only nominally indicated in the differential between the top and bottom collar speeds. This range must be related to the specific collateral to gauge the strength of the protection provided. The type of collateral—the agency, the differential between current mortgage rates and the mortgage rates on the underlying loans, seasoning of the loans, and the degree to which the pools have prepaid in the past—determines how quickly or slowly the collateral will prepay in different interest rate scenarios. The collars simply define a range of prepayment speeds over which the PAC payments will not vary. When prepayments fall outside the collars, payments to the PAC holders may be either delayed or accelerated (the collars could be broken temporarily without affecting the payment schedule). A given set of PAC collars will provide stronger or weaker protection, depending on the collateral. For example, a top collar of 300% PSA provides greater call protection should interest rates decline if the collateral is a current coupon than if it is a premium coupon. Similarly, a bottom collar of 100% PSA provides better protection from extension if the collateral is a conventional pass-through than if it is a GNMA.

All else being equal, PAC bonds backed by premium-coupon collateral do exhibit greater average life variability than PACs backed by current-coupon collateral. This is illustrated by the comparison in Exhibit 27–3, which displays the average lives at different prepayment speeds of two series of PACs, one backed by FNMA 9s (9.77% weighted average coupon [WAC], 349-month weighted average remaining term [WART]) and scheduled at 85% to 300% PSA, the other backed by FNMA 10 1/2s (11.13% WAC and 347-month WART) and scheduled at 95% to 350% PSA.2 (Similar comparisons could be made in the case of discount- and current-coupon collateral, but are omitted from this discussion for the sake of brevity.)

EXHIBIT 27–3
Impact of Collateral Coupon on the Average Life Variability of PAC Bonds

image

Investors recognize the stability lent by discount collateral and will accept a tighter spread for structures backed by discounts when they anticipate prepayments will accelerate. Similarly, bearish investors require wider spreads for PACs backed by premium collateral than if backed by current-coupon collateral. In markets characterized by bearish sentiment, the relative values can be reversed and investors will pay up for premium coupons and premium-coupon collateral while demanding a concession for discount collateral.

INTERACTION OF COLLARS AND COLLATERAL

Before further examining the contribution of the collar to the PAC’s value, it is useful to review the mechanics of defining a collar and creating a payment schedule. Visualizing the cash flows is also helpful in seeking to understand the behavior of the structure in different scenarios. The PAC schedule is defined by projecting the principal payments from the given collateral at a high and a low constant prepayment speed. Those payment amounts that can be satisfied by both sets of projected principal payments (that is, the smaller of the two amounts generated for each date) make up the schedule. This is depicted graphically in Exhibit 27–4. The example depicted was structured from $500 million FNMA 9s with a 9.77% WAC and a 349-month WART3 on the underlying mortgages, assuming collars of 85% to 300% PSA and a 2-year lockout (the period before the first scheduled principal payment).

EXHIBIT 27–4
Principal Cash Flows at 300% and 85% PSA—FNMA 9% Collateral

image

The faster speed results in a cash-flow pattern with the bulk of the principal thrown off during the first 5 to 7 years of the issue’s life. The slower speed produces a more level set of smaller cash flows extending to the final maturity of the collateral. The intersection of these two sets of payments forms the schedule. In the example, the bottom collar determines the payment amounts during the first 100 or so months of the schedule, and the top collar the amounts in the remaining months.

The graph of any speed between those of the top and bottom collars also would contain the area of the schedule below it. The graph of any speed faster than the top collar would bunch more principal in the first years and truncate the tail of the schedule in front of the point where the top and bottom collars intersect. Similarly, the graph of any speeds slower than the bottom collar would reduce the size of paydowns in the front years. At very slow speeds, most of the principal payments are pushed into the back years.

Once the PAC schedule is defined at the given collars, the PAC schedule may be further divided into classes with different average lives. An individual bond carved from a given schedule may have more prepayment protection than the scheduling collars would indicate. The example in Exhibit 27–5 is split into seven tranches. Some of the earlier bonds in the structure have higher effective top collars. An effective collar is the highest (or lowest) constant prepayment speed that would satisfy the entire tranche’s payments. For instance, many speeds faster than 300% PSA will contain the first tranche in this example. The fastest speed containing all of the first tranche is that bond’s effective upper collar. Effective top collars are shown in Exhibit 27–5 for the third and fourth tranches. These effective collars are approximately 370% and 330% PSA, respectively. It should be apparent as well that earlier tranches have still-higher effective collars.

EXHIBIT 27–5
Effective Top PAC Prepayment Collars—FNMA 9% Collateral

image

Likewise, the lower collars used to structure the PACs in Exhibits 27–4 and 27–5 do not fully indicate the degree of extension protection that the longer average life bonds, tranches 6 and 7, actually possess. Speeds below 85% PSA throw off principal too slowly to satisfy the scheduled payments in the early PAC tranches, but they provide more cash than needed to meet payments after about month 100, where the upper collar binds the schedule. For example, the effective bottom collar on tranche 7 is 55% PSA. Since tranche 6 begins to pay at the point where the top and bottom collars intersect, its effective collar is the same as the structuring collar.

When PACs first became popular, a wide variety of collar ranges and levels were used to generate the schedules, and even as late as the first quarter of 1989, it was not uncommon still to see some variety in collars on new CMO issues backed by similar collateral. However, the smorgasbord of prepayment collars largely has given way to increasing standardization of the speeds at which PAC schedules are created.

The standard speeds used to create PAC schedules may represent the market’s aggregate opinion of what constitutes a “good” collar; however, investors should (1) identify the effective collars for a bond and (2) translate the PSA collars into interest rate collars by using an econometric prepayment model to determine how much interest rates must shift to break the collars. Once the collar speeds are explicitly linked to interest rate shifts, investors can determine if the collar does deliver “good” protection over the scenarios appropriate to the investor’s outlook and portfolio.

New-issue PACs may be marketed either at the collars used to create the schedule or at their effective collars, although the trend has been to advertise effective collars. Unless the effective collar is known, the extent of the PACs protection against shifts in interest rates cannot be determined. Investors, therefore, should insist on this information, as well as analysis of the bond’s performance outside the effective collars.4 By the same token, investors cannot assume that a new-issue short-term average life PAC with a very high effective collar is better than a comparable average life PAC from a series with only the structuring collars indicated.5

With greater standardization of structuring speeds, the PAC market has developed a two-tiered structure, with standard PACs in the first tier and PACs with weaker collars in the second, trading significantly cheaper than “good” PACs. In mid 1989, issuers began to issue both types of PACs from the same transaction by layering a second set of PACs with narrower collars over the first. These two-layered PACs are, in essence, support bonds with schedules. They are discussed at length in Chapter 29.

In the 1991–1994 rally, the practice of carving “super-PACs” out of the first priority PAC schedule became widespread. The PACs which had second priority to the super-PACs still had standard new-issue PAC collars, but they were subject to much greater average life variability outside the collar speeds. As a result, they tended to trade behind standard PACs. As that “prepayment emergency” continued, collars became less important to astute investors who instead scrutinized bond performance across a spectrum of prepayment scenarios.

PAC COLLAR DRIFT

Many investors do not understand that the PAC collars are not fixed for the life of the tranche, but instead change over time with the actual prepayment experience of the collateral. This is evident from the number of investors who seek to evaluate trades in the secondary PAC market by looking at the collars advertised at issue (which could be either structuring or effective collars). Instead, off-the-run PACs should be evaluated by looking at the current effective collars. In most cases, they can be determined by the security dealers or third-party data services that have modeled the structure and know the current bond and collateral balances.

The effective collars simply express the highest and lowest constant prepayment speeds at which the given collateral can continue to meet the scheduled prepayments. Unless the collateral prepays at precisely the collar speed (and then it can match only one, the upper or the lower collar speed), it will have a different balance than was projected when the schedule was defined. Likewise, the amount of support bonds (and proportion of support bonds to PACs) will be different. Prepayment rates below the top collar cause the collar to shift upward over time (as there is a greater amount outstanding than anticipated), whereas the effective collar is lowered if prepayments are higher. Likewise, prepayments above the bottom collar cause it to rise. Short- and intermediate-term PACs are affected differently by speeds below the collar than are long-term PACs at the end or tail of the schedule. Very slow speeds cause the collar to drift up in the earlier tranches, while speeds below the bottom collar can improve the extension protection of long PACs. Prepayment rates somewhere between the top and bottom collars—historically, the common occurrence—cause the effective prepayment protection to widen as the top collar changes more quickly than the bottom collar (unless prepayments are very near the top collar).

WHEN THE PAC BREAKS

Breaking a PACs schedule and causing it to be partially called or extended are not necessarily negative events for the bond’s economic performance. The bond’s coupon relative to market yields (or the bond’s price relative to parity6) determines the effect prepayments outside the collar have on the realized yield or total return. A PAC with a discount coupon may benefit if the upper collar is broken, returning principal at par earlier than anticipated at pricing. Similarly, the extension caused by breaking the lower collar adversely affects the performance of a discount-coupon PAC. In the case of a premium-coupon PAC, it benefits performance when the bottom collar is broken, since the principal remains outstanding longer than anticipated, earning additional coupon interest. And of course, the higher the coupon, the worse the effect of breaking the top collar.

The shape of the yield-curve is also a consideration. Extension can be costly in a steep yield-curve; discounting a bond’s cash flows at a sharply higher yield can offset any value additional coupon income might have. Likewise, the capital gain from rolling down a steep curve can offset the loss of coupon income.

As collars deteriorate and approach a point where they may be broken, the market pays increasing attention to the average life profile. A PAC may no longer have measurable effective collars but may still have a stable average life over a reasonable range of prepayment rates.

WINDOWS

A PAC “window” is the interval over which scheduled principal payments are made to the bondholder. As long as prepayment speeds remain at a constant speed within the upper and lower protection bands, or PAC collars, the dates of the first and last principal payments (and equivalently, the length of the repayment period) are certain. PAC buyers in general prefer tighter windows. To some extent, this preference reflects the practicalities of managing their portfolios. The match between a single liability and a single asset is easier to conceptualize when payments are concentrated over a short period. They also are easier to convert to floating-rate assets with swaps. A shorter window also means fewer and larger (relative to investment amount) repayments. Tighter windows producing a more bulletlike pay down are, conceptually, better substitutes for corporates. The chief benefit of a tight window, however, is the superior roll down the yield-curve it provides as it ages. Ideally the average life declines by a year for every year the bond is outstanding (assuming a stable yield and prepayment scenario).

Generally, a tighter window is a greater consideration the longer the average life of the bond. Shorter-term PACs inherently have shorter windows, while the desire for tight windows in the 20-year sector is difficult to satisfy, owing to the “tail-ish” nature of the cash flows in the later years of the transaction. That is, the principal payments scheduled for the later years are relatively small in any month. This is easily confirmed by glancing at Exhibit 27–4. Carving the “tail” into shorter windows would result in more classes with odd, less marketable average lives.

A “good” window in the 3- and 4-year sectors is 12 to 15 months long. The 5- and 7-year buyers prefer a window of 18 to 24 months; 10-year buyers prefer paydowns over a 24- to 30-month period.

Some investors claim to prefer tight windows because they expect superior average life stability. For example, they may believe a tighter window has less average life variability when prepayments are outside the PAC collars. This hunch is disproved by experimenting with different window lengths in otherwise identical PACs (same collateral, same average lives). Exhibit 27–6 shows two such experiments manipulating the windows of the 5- and 10-year PACs in a structure containing a complete series of PACs with average lives from 2 to 20 years, backed by $500 million FNMA 9s (9.77% WAC, 349-month WART) and protected between 85% and 300% PSA. Even at fairly extreme prepayment speeds (0% and 600% PSA, for instance), there is little difference in the average lives of otherwise comparable PACs with different windows. All else equal, as a result of their slightly wider spreads, PACs with average or wide windows should outperform those with tight windows. This result suggests that investors who do not require a bulletlike repayment of their investment, but rather can accommodate greater payment dispersion, should not discriminate between window sizes, particularly in a flatter yield-curve environment. Investors who can adapt to a longer paydown period by such means as modifying their cash management procedures, adopting more sophisticated techniques for modeling and managing their asset/liability positions, initiating or other procedural changes, would also be able to take advantage of the relative cheapness of wide windows.

EXHIBIT 27–6
Impact of Window Size on the Average Life Variability of PAC Bonds

image

There is an exception. Tighter windows may improve performance when the PAC is a current pay bond. In this case, a shorter window reduces the likelihood that prepayments can accelerate or decelerate to levels outside the bands before it is fully retired according to schedule.

LOCKOUT

PAC bonds can be “locked out” in two ways. First, the entire PAC schedule can be locked out by shifting a portion of the PAC schedule at the very front to the support bond classes; resultant PAC bonds are locked out for the period over which those principal payments are made instead to support bonds. Typically this lockout could extend from the first 12 to 24 months of the issue’s life. The objective of the lockout is to stabilize the early support bond class. This is achieved by paying to the support bonds the principal cash flows that in effect have the highest effective collars (that is, they will be realized across a very wide range of prepays). These cash flows, which have a high degree of certainty, are used during the lockout to pay the support bonds. Some market participants, however, speak of the lockout as a PAC bond characteristic or feature; they may be viewing the lockout as a device for narrowing the PAC window. Others may perceive the lockout as somehow detrimental to the PACs in the structure, perhaps assuming that the PACs are somehow hurt to the extent the support bonds are helped.

Whether any of these assumptions are valid should be apparent from the example in Exhibit 27–7. The table contrasts the volatilities of 3-, 7-, and 10-year average life bonds from a structure without any lockout with those from one with a 2-year lockout. Both structures are backed by the same collateral, FNMA 9s, with a 9.77% WAC and 349-month WART, and use the maximum PAC schedule consistent with collars of 85% to 300% PSA, and, as appropriate, a lockout. The sizes of the bonds compared have been adjusted to match their average lives within two decimal places. (The 20-year bonds are not included in the comparison because their average lives were too different after matching the earlier bonds.)

EXHIBIT 27–7
Impact of Lockout Feature on the Average Life Variability of PAC Bonds

image

The lockout, according to the exhibit, benefits the PAC bonds by reducing both call and extension risk. Earlier bonds benefit more than later bonds, with the 10-year classes displaying only marginal reductions in average life volatility. Two effects are at work here. First, the lockout reduces the size of the schedule by removing all payments in the first two years. This results automatically in a smaller amount of PAC bonds relative to the support bonds; conversely, more support bonds protect the remaining PAC schedule. Since they contain those cash flows from the collateral with the lowest degree of call risk, the support bonds are much less vulnerable to call risk, and even at very high prepayment speeds a larger proportion of support bonds remain outstanding to shelter the PAC bonds than would otherwise have been the case. At the same time, these principal amounts are no longer bound to a schedule, meaning that later scheduled payments have a better likelihood of being paid on schedule in event of speeds below the bottom collar.

The second form of lockout is a natural form of call protection common to all types of CMO structures. It refers to the bonds (or principal amount of bonds) that will pay down, in the worst case, before the specific bond begins to pay principal. In a simple sequential structure, all but the shortest average life bond enjoy some lockout while earlier bonds in the structure pay down. PACs enjoy explicit call protection from the support bonds as well as inherent call protection from any earlier bonds in the schedule. In other words, the longer the average life, the better the bond’s inherent call protection. The impact of lockout is particularly apparent when we examine the effect of average life on option costs later in this article.

IS THERE A Z IN THE DEAL?

A Z, or accrual, bond (“Z” standing for zero coupon) is a type of CMO bond structure that pays no interest until it begins to pay principal. Until that time, the interest payments are accrued at the coupon rate and added to the principal amount outstanding. A Z-bond is most typically included in a CMO structure as the last bond class to be retired. Of course, the underlying collateral continues to pay coupon interest; the portion that would have gone to the Z-bond holders, had it been structured as a coupon-paying bond, is used instead to retire the earlier classes. In effect, then, the presence of a Z-bond permits CMO structurers to increase the size of the earlier classes, since the “accrual” amounts are additional to the projected principal payments from the collateral. More pertinently, the “accrual” helps to stabilize the earlier bonds, since a portion of the cash-flow used to retire them is not directly determined by the level of prepayments.

The interaction of Z-bonds with earlier classes is discussed in detail in Chapter 28. Readers who are unfamiliar with the Z-bond structure, or are interested in more complex manifestations of the Z structure, should refer to that chapter. The objective of this discussion is to examine the value, if any, that a long-term Z-bond contributes to PACs. A related issue also is explored, namely the possibility that accrual from the Z-bond is used to stabilize earlier support bonds and not the PAC bonds, or equivalently, to pay down stated-maturity-type bonds. Such a mechanism might not be explicitly disclosed when the bonds from the structure are traded, and may only be indicated in the prospectus or by careful review of the entire issue.

Two CMO structures were modeled to examine more closely the impact of a Z-bond on their performance and relative value, again backed by the same FNMA 9% collateral, and containing the maximum principal amount of PACs given collars of 85% to 300% PSA and the relevant accrual mechanism. The schedule has been divided into a series of bonds with nominal average lives of 2-, 3-, 4-, 5-, 7-, 10-, and 20-years, the average lives of all but the 20-year matching to within two decimal places (as in the lockout example, because the schedules differ significantly in amount, the 20-year bonds are not comparable).

Both structures contain 20-year Z-bonds. The first passes accrual to PAC and earlier support bonds alike, the second to earlier support bonds only. The average lives of the two PAC series at various prepayment rates are displayed in Exhibit 27–8.

EXHIBIT 27–8
Impact of Z-Bond Accrual on the Average Life Variability of PAC Bonds

image

Investors should note the difference in average life performance between the PACs supported by a Z-bond and the PACs whose support bonds alone are supported by the Z-bond. The Z-bond stabilizes the PAC bonds when prepayments break the lower collar. This makes sense—accrual cash-flow is generated as long as the Z-bond is outstanding. On the other hand, PAC bonds shorten more sharply in the structure with the Z-bond when the upper collar is broken, because a smaller proportion of support bonds is outstanding at any time to cushion PACs from high rates of prepayments. In fact, the faster the prepayments, the smaller the principal balance of the Z-bond when it begins to absorb excess cash, and the quicker it can be extinguished. This example should warn investors not to assume, however, that because a Z-bond is present in the structure, that the PACs will have less extension risk. There is no rule in the marketplace requiring issuers to pay accrual to PAC bonds. In some market environments, diverting accrual to the support bonds can make them more marketable, and a CMO arbitrage more viable. If this is the case, then issuers will structure their CMOs accordingly. The moral of the story: ask for the priorities in detail and read the prospectus.

Readers also may have noticed that the average life profile of the PACs that receive no accrual in this example is identical to that of the PACs backed by the same collateral in a structure with no Z-bond. (See Exhibit 27–3.) In other words, for PAC buyers, diverting accrual to the support bonds is the same as not including a Z-bond in the structure at all. This makes sense: The same aggregate amount of support bonds is available to support the PAC bonds, with the only difference being that the weight of support bond principal payments is shifted forward in time since the payments to the non-Z support bonds consist in part of interest accrued by the Z-bond.

EFFECT OF JUMP-Z AND VADM STRUCTURES ON PAC BONDS

In 1990, structurers began to create a special bond class from the accrual thrown off by a Z-bond. Variously called thrift liquidity bonds, VADMs (very accurately determined maturity), or SMAT (stated maturity), these bonds first appealed to savings and loan institutions, who are required by regulation to maintain a portion of their assets in very high quality, short-term investments, and who, accordingly, are willing to pay a premium for instruments that qualify and offer yields more attractive than those of, for example, government issues. These bonds appeal to investors such as commercial banks, who are sensitive to the extension risk associated with rising yield environments, so that VADMs with final maturities greater than two years have been issued in growing amounts. The size of a VADM class is determined by the amount of accrual (or accrual and principal) thrown off at a zero-prepayment rate up to the desired final maturity; obviously, faster prepayments will shorten the maturity, but no event can lengthen it. The presence of a VADM has the same effect on the PAC performance as diverting the accrual to the support bond classes has in the preceding example.

In a handful of issues, structurers designed the support bond Z-bonds to convert to a “payer” if certain trigger conditions occur, changing their payment priority among support bonds from last to current. There are a variety of ways to make a Z “jump”; the chief ones are described in Chapter 28. Once the Z-bond converts to the current pay bond and begins to pay coupon interest, its support is no longer available to the PAC classes. The effect on PAC performance is generally the same as in the basic example above. However, the degree to which the PACs lose extension protection is moderated by the amount of time it takes to trigger the conversion, and by whether the jump is a temporary response to some condition (such as a specified prepayment threshold) or permanent (a “sticky” Z). This structuring strategy was more common in the late 1980s and hasn’t been seen in recent market environments.

PRIORITY TO RECEIVE EXCESS CASH FLOWS

The PAC schedule is protected by the existence of support bond classes. The mechanism is simple: In any period the support bonds absorb all principal in excess of the scheduled payments, and any current-paying PACs have first claim on all principal received. This protection ceases when all the support bond classes have been fully retired, an event that occurs if the collateral consistently pays at speeds above the top collar. Once the support bonds are retired, any principal is distributed to the outstanding PACs, according to priorities defined for the particular CMO issue. Frequently, these priorities pay excess principal to the outstanding PACs in order of final maturity, but this is not always the case. A wholesale examination of CMO prospectuses will unearth numerous examples of structures that paid excess in the reverse of maturity order or otherwise insulated some classes at the expense of others. The impact of such schemes has, as would be expected, a significant effect on the average life volatility of the various PACs.

A simple example contrasting two priority schemes is shown in Exhibit 27–9. This exhibit compares the average lives of two sets of PACs at various prepayment speeds—one receiving excess principal in the maturity order, the other in the reverse maturity order. As the exhibit indicates, reversing the order in which PACs are subjected to prepayments above the top collar drastically alters the average life performance of the PACs, if prepayments are outside the collars. As would be expected, shorter PACs benefit at the expense of the longer. When the excess is paid in reverse, the short- and intermediate-term PACs are more significantly stable; when prepayments increase, longer PACs, with 10- and 20-year average lives, shorten up much more significantly.

EXHIBIT 27–9
Impact of Excess Payment Order on the Average Life Variability of PAC Bonds

image

THE OPTION COSTS OF PAC FEATURES

Most participants in the PAC market evaluate individual bonds by examining their average life and yield over various constant prepayment scenarios outside the collars, a technique similar to the one used in this chapter to analyze the various PAC features. The procedure has recognized disadvantages, some of which can be reduced to the complaint that they use a constant prepayment assumption. Using such tools, investors can devise investment criteria for PACs such as “I will buy 10-year PACs with 4-year windows at 120 off if they don’t shorten to less than 8 years average life given an instantaneous 200-basis point (bp) drop in yields.” Implicitly, they are using these tools to measure and value the prepayment options embedded in their PAC bonds. However savvy and tough-minded the criteria sound, they are at bottom purely subjective guesses about how much the random exercise of those options will impair or help investment results.

It is possible to measure the impact of prepayment risk on the yields earned by PACs by employing option pricing models. These models generate spreads, durations (price sensitivity), and convexities (the sensitivity of duration to yield changes) that are explicitly adjusted to account for expected prepayments on a large sample of possible interest rate paths over the life of the PAC. In particular, these models derive an average cost of the prepayment options in the PAC, and determine the expected reduction in total spread caused by interest rate volatility.7 The model, in simple terms, summarizes hundreds more scenarios than investors can digest looking at price-yield tables. The scenarios, moreover, are more realistic in that they permit interest rates to move randomly at each point along the path.

Option-adjusted spreads (OASs) vary with market conditions. For this reason, a discussion of current OASs is inappropriate here. However, the option costs derived from the analysis are not as sensitive to current market yield levels (although they will vary somewhat with changes in the level of implied volatility), and can be discussed here without becoming hopelessly stale with the next rally or correction in fixed income markets.

As expected, PACs demonstrate low option costs. Option costs in the current-coupon-backed structures discussed in this chapter generally ranged from 0 to 20 bps for bonds with 3- to 20-year average lives. By comparison, the collateral (FNMA 9s) had 35 bps of option cost. (Bear in mind that absolute measures of option costs, yields, and so forth are calibrated to a specific option model. PAC and other CMO option-adjusted spreads and costs should always be benchmarked to those demonstrated by the collateral in the same model.) The PAC bonds backed by premium collateral (FNMA 10 l/2s) had demonstrably higher option costs, ranging from 3 to over 40 bps. The collateral had an option cost of 68 bps.

The general pattern revealed by the option-pricing model is illustrated in Exhibit 27–10. The option costs are calculated for the series, described in earlier sections, of 2-, 3-, 4-, 5-, 7-, and 20-year average life-PACs backed by FNMA 9s. (See Exhibit 27–3 for the average life profile of these bonds.) The patterns displayed by option costs for premium-backed PACs are shown in Exhibit 27–11. The option costs for these series are analyzed in three scenarios: assuming that the interest rate environment (of January 23, 1990) remains constant and assuming instantaneous parallel shifts in interest rates of up and down 100 bps. Assuming a constant option-adjusted spread, the shifts in the up and down cases are typically large enough to give the tranches, which currently are priced close to par, a discount or premium price. In this way it is possible to draw some conclusions about the sensitivity of the structure to interest rate shocks. Ranging from 3 to over 30 bps, the option costs demonstrated by the current-coupon-backed PACs in the bullish scenario are consistent, as well, with the general magnitude of option costs observed in premium-coupon-backed PACs.

EXHIBIT 27–10
Sensitivity of Option Costs in PACs* to Interest Rate Shifts (Current-Coupon Conventional Collateral)

image

EXHIBIT 27–11
Sensitivity of Option Costs in PACs* to Interest Rate Shifts (Premium-Coupon Conventional Collateral)

image

As expected, the option costs rise with the average life of the PAC. This result is expected if only because the probability an option will be exercised is greater the later its expiration date. Comparing the current and bearish cases, the increase in extension risk as interest rates rise results in a slightly higher option cost for 2- and 3-year PACs. Of greater interest is the fact that, going from the 10-year PAC to the 20-year, option costs either decline or increase at a slower rate in every interest rate case. This result holds for the option costs in a similar set of PACs depicted in Exhibit 27–11, backed in this instance by premium coupon collateral (the average life profile is shown in Exhibit 27–3). The fact that option costs tend to peak with 10-year PACs rather than 20-year PACs should take some market participants by surprise, and suggest to others that the typical incremental spread required by 20-year PAC buyers over 10-year PACs represents extra value. The result is intuitively appealing as well. It reflects the long lockout imparted by comparisons and shorter PACs in the schedule and the limited room for extension. In addition, the prepayment model used to project prepayments over various interest rate paths correctly reflects the tendency of mortgage pools to “burn out” following sustained periods of fast prepayment rates. As a result, prepayments tend to slow down in the later years of the pool’s life.

When the effect of a lockout was examined using average life profiles (Exhibit 27–7), the benefits to the PACs were more pronounced the earlier the bond came in the series. By explicitly valuing the option costs for the same example, a similar effect is observed, as shown in Exhibit 27–12. However, the 7- and 10-year tranches receive no benefit from the lockout on an option-cost basis, whereas their average lives are observed to lengthen less sharply in extremely slow prepayment rate environments. The exhibit also indicates a benefit to the 20-year PAC, except when interest rates drop and the collateral becomes a premium security and subject to faster prepayments. The 20-year PACs were not compared in Exhibit 27–6 since, given a different amount in the schedules, it was not possible to match the average lives of all four bonds (the 20-year in the lockout structure had an average life close to 16 years). This difference also may account for the lower option cost imputed to the 20-year in the lockout structure.

EXHIBIT 27–12
Sensitivity of Option Costs in PACs* to Interest Rate Shifts in Structures with and without a Lockout

image

The impact of various window lengths on 5- and 10-year PACs also was examined, with no increase or diminution of option costs observed, except for the 5-year bonds in the bullish case. Even then, only slight differences, at best 3 bps, were manifested. This reiterates the conclusion, stated earlier, that the length of the window does not significantly affect the average life stability of the bond outside the collars.

Some security analysts and investors resist this result. They believe, for instance, that a tight window lowers the likelihood of breaking the collars during the paydown period. Therefore, they reason, the prepayment options that they effectively hold should be less costly. The time value of the options, however, includes the period prior to the first payment, because the protection implicit in the collars can be damaged or enhanced by prepayment experience in earlier months or years of the structure’s life. More than one full interest rate and housing industry cycle can occur before a single principal payment is made to a 5- or 7-year PAC, with periods of slower prepayments tending to improve a PAC’s call protection and faster prepayments tending to erode it. If prepayments violate the lower collar when earlier PACs are paying, both the call and extension protection of the later PACs can actually improve! When many such possibilities are simulated, the net effect should be small or negligible.8

The option costs for two series of PACs in Z-bond structures, one funded by the Z-bond and one not, are displayed in Exhibit 27–13. The fact that the Z-bond helps reduce the extension risk in the PACs but exposes them to additional call risk is illustrated. The PACs paid down with accrual tend to have higher option costs, except in the bearish case, where prepayments are less likely to retire the Z-bond before the PACs have been paid. (The relationship appears to weaken for 20-year bonds, but this most likely reflects the fact that the 20-year bonds have very different average lives as a result of matching the earlier bonds. The 20-year PAC paid by accrual actually has an average life of over 26 years.)

EXHIBIT 27–13
Sensitivity of Option Costs in PACs* to Interest Rate Shifts When Z-Bond Funds Supports Only and When Z-Bond Funds Both PACs and Supports*

image

Reversing the priorities when the top collar is broken, so that the last PAC in the schedule is the first to receive excess principal after the support bonds are retired, has a very large effect on the option costs in the 7-, 10-, and 20-year PACs. This effect is apparent in Exhibit 27–14. The higher the priority, the greater the call risk and the higher the option costs. The impact is, as expected, accentuated as declining yields elevate the risk of prepayment. By contrast, shifting the call risk to the later tranches strips most of the already low option costs from the earlier tranches.

EXHIBIT 27–14
Sensitivity of Option Costs in PACs to Interest Rate Shifts When Excess Cash Is Paid to PACs in Sequential and Reverse Orders

image

KEY POINTS

• The scheduled payments of PACs, protected over a wide range of possible prepayment speeds, appeal primarily to insurance companies and other buy-and-hold investors who are matching specific liabilities. The liquidity, yields, and wide diversity of features in the market also attract growing numbers of active bond managers.

• Misconceptions about the value of certain PAC features can create a number of opportunities for investors in both groups. Most notably, many premium coupon-backed PACs may be undervalued when their current effective collars are taken into account; the length of the PAC window does not contribute to economic value; lockouts benefit short-term average life PACs as well as support bonds; and Z-bonds can protect earlier PAC bonds from extension risk.

• Investors who fail to stress-test PAC investments may overlook the fact that effective collars are significantly different from those stated at issue. Any kink or deviation from sequential order in the prioritization of excess cash-flow to the PACs after support bond classes are retired must be detected, as it could drastically affect the performance of the longer-term PACs. Although the impact on value is less dramatic, investors should also determine whether a long-term Z-bond pays support bonds or PACs in the structure.

• PACs are normally evaluated by examining the yields and average lives of PAC investments over a variety of prepayment scenarios. Such analyses must be carried a step further and linked to possible interest rate scenarios. The soundest way to do this is to employ a prepayment model that explicitly recognizes the determinants of prepayment behavior. Such an analysis can be supplemented and its insights extended by employing option-based pricing methods.

..................Content has been hidden....................

You can't read the all page of ebook, please click here login for view all page.
Reset
54.198.108.174