4.3 PRODUCT MECHANICS AND APPLICATIONS
4.3.1 Mortgage-backed securities
r
creates an efficient mechanism by which to purchase a desired asset – administrative savings result when an investor purchases a tranche of an entire portfolio of assets through a single transaction;r
permits companies and other originators of future cash flows to monetize their stream of forward earnings for use in a current period;r
allows originators to transfer certain cash flow uncertainties (e.g. prepayments, credit de- faults) and risk exposures to investors, at a price;r
transforms otherwise illiquid mortgage and receivable assets into a more marketable form, adding greatly to financial sector liquidity – this ultimately helps lower issuer funding costs and creates more attractive and secure opportunities for investors.Since risk/return characteristics of MBS and ABS are tailored via different tranches, they are suitable for a wide spectrum of investors, primarily from the institutional markets. Accordingly, purchasers of MBS and ABS include financial institutions, insurers, pension, mutual, and investment funds, and corporates. For instance, investors preferring little prepayment risk may be attracted to planned amortization class bonds; those seeking a significant amount of exposure to interest rate movements may purchase interest-only/principal-only tranches, and so forth.
for increasing, balloon, or bullet payments of principal). However, interest rates may decline over time, inducing the borrower to repay the original mortgage and refinance through a new one with a lower interest rate (evidence suggests that refinancing becomes attractive when rates have declined by at least 100 bps below the original rate).4 Alternatively, the borrower may choose to sell the underlying property, repaying the original mortgage in the process.5Or, the borrower may wish to make greater principal payments during a given month(s) in order to build equity more rapidly. In the worst case, the borrower may be unable to continue paying the mortgage, and declare a default. Each one of these payment/prepayment actions creates cash flow uncertainties that make the valuation of MBS a nontrivial task. Note that adjustable rate mortgages (ARMs) can be impacted by the same variables, as well as the changing value of the interest rate component (which may be readjusted on a monthly, quarterly, semi-annual, or annual basis).6From an MBS investor’s perspective, mortgage cash flows can be viewed as the sum of projected monthly interest less servicing fees, projected monthly scheduled principal repayments, and projected monthly additional principal prepayments.
Most prepayment activity is due to home sales when rates are stable. As rates rise, refinanc- ing opportunities disappear and housing turnover slows, meaning that prepayments decline.
Conversely, as rates fall, refinancing commences and housing turnover increases, leading to prepayment acceleration; we consider this at greater length below.
Refinancing, fuelled primarily by declining interest rates,7remains the single largest driver of prepayments, and is the central focus of any prepayment model. This means, of course, that projected prepayment experience is extremely dependent on assumptions regarding future interest rates. Refinancing can be considered the equivalent of a call option on interest rates owned by the borrower; the option cannot, unfortunately, be modeled in a standard option pricing framework, due to the behavioral inefficiencies that characterize the housing market.
That said, there is sufficient empirical evidence to suggest how refinancing patterns arise during a given cycle, and these results can be incorporated into a model. For instance, most cycles feature a burnout period, meaning that after a burst of refinancings, further refinancing activity tends to slow; each marginal decrease in rates yields less refinancing activity. Friction costs and barriers have declined in recent decades, making the refinancing process simpler, cheaper, and more efficient; the advent of online refinancing technologies, for instance, has allowed a greater number of borrowers to refinance during a particular cycle, meaning that the number of
4When refinancing opportunities first appear, prepayment activity accelerates until a critical mass of activity is done; this phase is known as the “burnout,” after which any further refinancing in response to a continued drop in rates is likely to be marginal.
5Housing turnover through home sales is a major driver of prepayments, historically averaging up to 10 % in a given year.
Housing turnover is, itself, a function of various economic variables, including interest rates, consumer confidence, new construction, seasonality, current and expected economic growth, and so forth.
6In fact, prepayments on ARMs tend to rise after the first six or twelve months as coupons reset (many resets eliminate the first period below-market “teaser” rate); the rising rates following the first period can lead some borrowers to refinance into new teaser rate ARMs (in fact, the number of new ARM-related products and the extremely competitive environment have led to a greater number of ARM refinancing opportunities – this, in turn, has led to a gradual increase in prepayment speeds in recent years). Once seasoned, ARMs settle in a range where prepayments are slightly above fixed-rate mortgages. When modeling prepayments on ARMs, some models distinguish between refinancing into new ARMs and new fixed rate mortgages; the nature of the borrower dictates the type of refinancing that is likely to occur: borrowers seeking to take advantage of teaser rates will refinance into new teaser ARMs, while those that prefer the stability of long-term fixed rates, but may have been unable to gain approval at a previous time, will refinance into standard fixed mortgage products. Other ARM-related products – such as convertible ARMs that change from floating to fixed rates after a specific period of time, and hybrid ARMs that convert from fixed to floating rate after five to ten years – tend to have unique prepayment characteristics of their own. Hybrids have proven especially popular in some markets, and growth has increased steadily as rates are often more competitive than standard 30 year fixed mortgages, and borrowers are able to defer the shock of migrating to a floating rate for up to five years. As a result of these benefits, borrowers often prefer to preserve rather than refinance their mortgages;
prepays are, accordingly, slower than on traditional ARMs.
7A smaller amount of refinancing occurs as a result of credit improvement, e.g. a borrower’s financial circumstances and credit standing improve, allowing migration from the subprime category to a standard credit category that commands a lower borrowing rate.
When this happens, refinancing is a logical option.
“lagging” refinancers has declined. Refinancing incentives8and mortgage product availability9 must also feature in the model. These can be supplemented by other, less tangible, forces – media coverage of the rate environment and refinancing opportunities, promotional campaigns sponsored by lending institutions, “psychological” rate barriers that might drive activity, and so forth. The micro and macro construction of the pool must also be considered. The market features many types of mortgages and rates, some of which promote refinancing action more readily and efficiently than others.10In addition, the diversity of borrowers in any given pool leads to a broad range of refinancing options. Slow refinancers will eventually comprise a larger proportion of a seasoned pool as the fast refinancers depart, which will again impact on prepayment speeds and valuation. Refinancings cannot, of course, be viewed in isolation:
even if rates are declining to the point where borrowers can achieve real savings, an associated deterioration in housing prices can impact behavior.
Housing sales, the second most significant driver of prepayments, and therefore a vital modeling input,11 can be impacted by overall market turnover, regional/local turnover, sea- soning, and lock-in/prepayment penalties.12Each of these variables, in turn, can be directly or indirectly influenced by other variables. For instance, overall and regional turnover may be affected by new housing starts and new home sales, economic growth, employment, taxes, interest rates, housing price inflation and propensity for “trading up.”13 Seasonal or cyclical effects must also be considered in the model.
Actual prepayment levels are determined by comparing principal cash flows received with those that are scheduled or expected; any difference between the two represents a prepayment, and generally is expressed in terms of the outstanding balance of the mortgage. In order to assess the potential impact of prepayment risk on mortgages – and by direct extension the impact on mortgage pools comprising an MBS – the industry has come to rely on certain simplifying assumptions designed to serve as a proxy for prepayment behavior. The most fundamental computation is based on the single monthly mortality (SMM) gauge, which is a monthly prepayment indicator that computes the fraction of a pool’s balance that prepays
8Refinancing incentives can be regarded, in simplified terms, as the spread differential between the WAC and the mortgage rate;
this, however, fails to account for the outstanding loan balance and the term of the loan. A more sophisticated approach compares the present value of all applicable inflows and outflows over the term of the loan:
R S= Bal*ro
rn
* 1−yt
1−zt
−(Bal*(1+RC))
whereR Sis refinancing savings,Balis the current balance of the loan,rois the original loan rate, rnis the new loan rate,RCis refinancing costs,yis 1/(1+ rn),zis 1/(1 +ro), andtis the remaining term of the loan.
9Models generally distinguish between ARM and fixed-rate mortgage products; the most sophisticated can model changes gener- ated when borrowers move from ARM to fixed-rate mortgages, or from ARMs to new ARMs. ARM pools generally are characterized by high mobility and fast seasoning, and are very sensitive to refinancing opportunities. Thus, if a pool is comprised solely of ARMs (a highly unlikely case, but useful for purposes of an example), the prepayment and valuation results will be very different than if a pool consists only of fixed rate mortgages.
10For instance, a 15-year mortgage requires larger monthly payments than a 30-year mortgage, but has a faster amortization profile and may feature a lower rate. The range of new mortgage products available in many markets gives borrowers considerable flexibility.
For example, borrowers can choose no-point or low point mortgages, cash-out refinancing loans (a form of home equity lending), and so forth.
11Though housing sales are an important driver of prepayments, historical evidence suggests that they tend not to create speeds in excess of 10 % PSA CPR in a given cycle. Refinancings, in contrast, generally are multiples of that figure.
12A lock-in provision or prepayment penalty provision on a loan can also be quantified to determine whether rates will have an impact on prepayment behavior. For instance, if the lock-in rate on a loan is below current market rates, a borrower has an incentive to keep the current property; if current market rates fall below the lock-in rate, the incentive begins to fade. The actual dollar impact can be analyzed by examining the present value impact of current versus expected market rates. A prepayment penalty provision requires that the dollar value of the penalty be compared with the cost savings that can be achieved in a refinancing scenario when market rates fall below the original loan rate; prepayment penalties are unusual in residential mortgages, but they are quite common in commercial mortgages.
13In some markets, activity is driven by the desire or need for existing homeowners to sell their original homes and purchase larger ones; if the trade-up occurs in the same locality, there may thus be purchases and sales impacting two sets of properties per household.
Figure 4.2 PSA CPR speeds
during the month. The constant prepayment rate (CPR) model serves as an annualized version of the SMM. The CPR assumes no difference in seasoning (i.e. the amount of time a loan has been outstanding), meaning that an assumed prepayment rate of 2 % applies to a loan that has been outstanding for six months or five years; this assumption is considered unrealistic in light of available historical data.
In the mid-1980s, the US Public Security Association (PSA) introduced a variation on the SMM/CPR approach to account for seasoning phenomena. The PSA, in examining historical data, determined that prepayment patterns change based on loan life: specifically, new loans have lower prepayment rates than seasoned ones. The PSA model adjusts the CPR by the age of the loan. The base case PSA model, known as 100 % CPR, assumes 0 % prepayment for new loans, 0.2 % prepayment for the first month, increasing by 0.2 % per month for the next 30 months, and then converting to a flat 6 % per year thereafter. This 100 % benchmark can then be increased (e.g. to 150 % or 200 %) to reflect faster prepayment scenarios, or decreased (e.g.
to 50 % or 75 %) to reflect slower prepayments. Figure 4.2 highlights various PSA CPR speeds.
It is common in MBS valuation to perform scenario analyses to determine interest rate sen- sitivity of the CPR and, by extension, the sensitivity of the WAM and yield. For instance, Fig- ure 4.3 illustrates a hypothetical CPR curve for changes in interest rates; this curve, which can be constructed from historical prepayment data, can then be used to generate possible MBS values.
Ultimately, an effective prepayment model must take account of all of the variables that can impact prepayments in order to provide an indication of possible future prepayment speeds, as noted in Figure 4.4, so that a specific MBS can be valued properly. In some instances, models are atomized to project each individual risk variable independently; these must, of course, be internally consistent so that the inputs and outputs among each independent submodel provide rational explanations of behavior (e.g. the effects of housing turnover from one submodel must be consistent with the effects of refinancing behavior from a second submodel). By creating this additional level of detail, an intermediary or investor can gain additional insight into future prepayments and security value.14Importantly, the efficacy of a model can be back-tested by running simulations against actual prepayment experience.
14The impact of different aspects of prepayment risk – such as refinancing, housing turnover, and so forth – can be ascertained, in part, by the use of partial durations, i.e. the change in the price of a security for a unit change in each of the specific variables through to impact on prepayments.
CPR (%)
Interest rates
High/increasing interest rates, slow/no refinancings Low/declining interest rates,
rapid refinancings
Figure 4.3 Effect of interest rate scenarios on CPR
Different types of MBS have different prepayment characteristics and sensitivities. Thus, in creating a workable model, attention must be given to the type of security being analyzed/traded;
there is no single model that can compute prepayments and valuations properly across multiple types of instruments without some adjustment for the specifics of the asset/pool. For instance, GNMA pools, which we discuss below, tend to feature higher LTVs than commercial pools.
High LTVs dampen prepayment speeds, leading to different values than commercial pools with lower LTVs. However, as GNMA loans season (and assuming housing market stability), LTVs drop and prepayments rise compared to commercial pools. Various other differences exist between other classes of MBS.
The modeling of prepayments must be viewed as a dynamic process, with parameters that change over time as market, borrower, and investor characteristics change. For instance, the
CPR %
Actual experience
Projected experience
Time
Model output x Model output 2 Model output 1
Figure 4.4 Prepayment projections from a model
introduction of new mortgage products, fundamental changes in interest rate structure, or differences in borrower behavior can all combine to permanently alter prepayments. A model that fails to take account of these changes will soon be of limited use.
Valuation
The prepayment risks described above are a fundamental input in the valuation process. Indeed, intermediaries and investors attempt to model prepayment behavior so that they can use the output to value an MBS more accurately. There are various ways of valuing MBS; some are elemental and simplistic but easy to implement, while others are computationally rigorous, though may ultimately prove to be more accurate. The potential value/return of an MBS can be analyzed through a simple static yield computation, or the more complex option adjusted spread (OAS) process we introduced in Chapter 3 (based on the concept that an MBS is an option-embedded security).
The static yield calculation is a cash flow yield derived from assumptions about prepayment speeds; it assumes that the investor will reinvest at the yield to maturity rate, and that the MBS is held until the final payoff date based on the prevailing prepayment assumption. The reinvestment assumption is critical, as prepayments (and thus reinvestments) occur every month or quarter; if cash flows differ from the assumption, then the cash flow yield will differ as well.
Under a static cash flow yield process, the yield spread of an MBS is equal to the difference between the cash flow yield and the yield to maturity of the risk free benchmark (e.g. a Treasury bond); this, however, fails to account for the term structure of the benchmark and expected interest rate volatility, which will impact prepayments and cash flows.
We extend the OAS concept discussed in the last chapter to the option-embedded MBS.
The OAS values an MBS by comparing the security to a mini-portfolio of zero coupon risk free benchmarks that have certain projected cash flows; the portfolio can be adjusted to re- flect the MBS default risk and cash flow uncertainty. The resulting spread makes the present value of projected cash flows from MBS, discounted at spot rate and spread, equal to the market price. The OAS process begins with the computation of a yield curve spread as a measure of the return of principal over multiple periods, with each MBS cash flow discounted by the appropriate risk free benchmark and spread using forward rates. However, the yield curve spread still assumes constant rates and cash flows, so it is only appropriate for current scenarios. In the second step, the OAS model computes the yield curve spread through a series of forward rates and then captures the effects of rate volatility to suggest a possible path of future rates.15Specifically, the OAS model creates random interest rate paths based on model parameters, and each path is used to project prepayment rates and MBS cash flows. Once computed, the values are discounted to generate an estimate of the average present value.16 The cost of the embedded option is simply the difference between the OAS and the yield curve spread, and can be interpreted as a measure of the investor’s cost of rate volatility.
Since OAS models are based on many assumptions, care is required in implementation and interpretation.
15In practice, most OAS models use an arbitrage-free term structure model calibrated to the current market. Models may be 1- or 2-factor, with or without mean reversion. Rate volatility must be incorporated properly, as short-term rates are more volatile than long-term rates.
16The model can, by extension, be used to compute option-adjusted duration and option adjusted convexity in order to determine the true price sensitivity of a particular security. Duration generates the percentage price change for anxbp parallel shift in the curve, assuming OAS is constant; securities that are sensitive to prepayments feature a duration that is lower than the standard modified duration of other fixed income assets. Convexity acts in a similar manner.