CHAPTER
TWENTY-FIVE
AGENCY MORTGAGE-BACKED SECURITIES

ANDREW DAVIDSON

President
Andrew Davidson & Co., Inc.

ANNE CHING, CFA

Senior Analyst
Andrew Davidson & Co., Inc.

EKNATH BELBASE, PH.D.

Senior Consultant
Andrew Davidson & Co., Inc.

Agency mortgage-backed securities (MBS) are a large and important asset class. They represent a core fixed income sector with substantial liquidity, and as reinforced by the financial crisis of 2007–2009, they have to date had essentially no credit risk and are preferred investments from the standpoint of capital requirements and regulatory treatment. Yet at the same time, they pose risks to investors that are often difficult to grasp and measure. This is due to the uncertainty of prepayments on the underlying mortgages, the complexity of borrower prepayment behavior, and the interplay between the credit risk of the underlying mortgages and prepayments on agency pools. In this chapter, we provide an overview of the distinguishing characteristics of Agency MBS, securities issued by government-sponsored enterprises (GSEs), and provide an analytic framework for evaluating the risk and relative value of these complex securities.

MORTGAGE LOANS

Residential mortgage loans are the building blocks of MBS, and as of year-end 2010, were the largest category of debt in the U.S. economy. In order to develop an understanding of how MBS work, one must first become familiar with the structure and characteristics of mortgages.

A mortgage is a loan that is secured by real property. A borrower or mortgagor is under legal obligation to repay the loan or risk forfeiture of the property backing the loan. A mortgage loan consists of two parts: the mortgage deed or deed of trust and the promissory note. The mortgage deed describes the real estate used as collateral against the repayment of the note. The promissory note represents the personal promise to repay the loan, which also stipulates the financial terms of the loan, including the rights of both the borrower and lender.

Mortgages are generally categorized by time to maturity, interest-rate type, amortization structure, and loan size. A fixed-rate mortgage, as the name implies, has a fixed-rate of interest for the length of the loan. Loan terms are typically 15 years or 30 years, with other terms (20, 25, and 40 years) much less common. The most common type of fixed-rate mortgage is a constant payment (also known as level-pay) fully amortizing mortgage. This means that the principal is fully amortized over the length of the loan and each monthly payment is the same dollar amount. In other words, the principal is repaid gradually over the term of the loan in contrast to U.S. Treasury securities and other bonds in which the principal is generally repaid in one lump sum at maturity; each constant payment is a changing mix of interest and principal, with the amount of interest decreasing over time and the amount that is principal rising, as shown in Exhibit 25–1.

EXHIBIT 25–1
Constant Payment Mortgage

image

Other types of amortization structure include balloon loans and interest-only period loans. A balloon loan requires a full repayment of remaining principal at a future date. For example, a 30-year fixed-rate loan with a balloon at 10 years requires all remaining owed principal to be repaid at month 120. An interest-only period loan allows the borrower to make smaller payments, consisting of only interest, for an initial period (such as 5 or 10 years), followed by larger payments after that date due to the shorter remaining amortization period. Finally, negative amortization mortgages allow an initial period in which borrowers are allowed to make payments less than the interest due, resulting in increasing loan balances.

Adjustable-rate mortgages (ARMs) have variable rates of interest. Lenders sometimes offer ARMs at below market “teaser” rates for an initial period of time. After the teaser rate period has expired, the coupon resets off a specified index. Typically, the index is tied to a comparable term U.S. Treasury bill or other market cost of funds such as the London Interbank Offer Rate (LIBOR) or Eleventh District Cost of Funds Index (COFI). An acceptable index from a regulator’s point of view is one that is not susceptible to market manipulation by the lending institution and is driven by market conditions. Typically the contract interest rate resets once a year subject to cap limitations. Periodic caps limit the rate increase at reset date, while lifetime caps limit the upper bounds of a coupon during the life of the loan. As a result, monthly payments adjust up or down as interest rates move.

Hybrid ARMs blend the features of both fixed-rate and adjustable-rate mortgages and began to be increasingly popular beginning around 2002. Hybrid ARMs are initially fixed-rate loans that convert into adjustable-rate loans after a predetermined period—typically 3, 5, 7, or 10 years—and then reset annually at a specified spread to an index over the remainder of a 30-year term. Other salient features of hybrid ARMs include cap structures and indices. Hybrid ARMs are subject to initial, periodic, and lifetime caps. The initial cap limits the amount the contract rate of interest can rise at the first reset date. The periodic cap limits the increase in the contract rate at each subsequent reset date. The lifetime cap establishes a ceiling on the amount the contract rate can increase over the life of the loan. The most common cap structures for hybrid ARMs are presented in Exhibit 25–2. In addition some 5/1 hybrids have a 2/2/5 cap structure.

EXHIBIT 25–2
Typical Hybrid ARM Cap Structures

image

Hybrid ARMs have been issued with a variety of indices, with the 1-year Constant Maturity Treasury (CMT) as the most prevalent index prior to 2001. There has been a recent trend toward LIBOR-indexed hybrids, which reflects liquidity concerns with respect to the disappearance of the 1-year Treasury bill. Additionally, LIBOR-based MBS provides a closer match between bank assets and liabilities, reducing basis risk between LIBOR-based liabilities and ARMs. In 2003, Fannie Mae introduced a LIBOR-indexed 5/1 hybrid program as a way to create more uniformity and liquidity in the market for LIBOR-based hybrid products.

While interest-only periods are available with fixed-rate loans, in the mid-2000s they became increasingly associated with hybrid loans. Meanwhile, balloon loans all but disappeared as hybrid loans became more popular. Finally, while negative amortization loans are theoretically possible with fixed-rate loans and a single amortization change date, negative amortization loans originated in the 2000s have tended to be adjustable-rate or hybrids, with the “Option-ARM” product gaining the most notoriety during the housing boom and aftermath of 2004 through 2010.

Another distinguishing feature of a mortgage loan is the absolute size of the loan. For example, as of 2011, a loan for a single-family home which is less than $417,000 is considered a conforming loan size, with other, higher limits for higher cost geographic areas. A loan whose size is greater than the conforming loan cut-off is called a jumbo loan. This limit has been adjusted annually based on national house price growth. Approximately 80% of mortgages in the United States meet the conforming loan size.

Exhibit 25–3 provides a list of various mortgage products available as of March 2004. Notice that the first two products are both 30-year fixed-rate mortgages. However, the contracted rates of interest differ because the first product requires two percentage points of the loan amount paid upfront by the borrower at the time of loan origination. The second 30-year loan does not require a payment of points. Points (sometimes referred to as discount points) are paid to the lender to reduce the overall interest rate of the loan. Borrowers who plan to live in a dwelling for a long period of time tend to pay points up front to lower their interest rate and thus monthly mortgage payment. Analysts consider the choice of borrowers to pay “points” to be predictive of the length of expected tenure.

EXHIBIT 25–3
Mortgage Products and Rates Available on March 5, 2004

image

Further down Exhibit 25–3 the interest rates of two fixed-rate jumbo mortgages (30- and 15-year) generally command higher interest rates than their conforming counterparts. This is both because jumbo loans tend to have more ruthless exercise of the borrowers option to refinance (more on this in our prepayment section) and because jumbo loans do not qualify for the agency loan programs, resulting in lowered liquidity for jumbo-backed securitizations. Over time, the spread between jumbo loans and agency-eligible loans has fluctuated, depending on factors such as the liquidity premium commanded by agency MBS, credit concerns and the general level of credit spreads, and the price of prepayment risk. The ARM products listed have lower interest rates than fixed-rate products because these mortgages are fully adjustable after the initial fixed period has expired, and as of the date shown in the table, the yield curve was relatively steep. Over time, the ARM/FRM difference primarily varies as a function of the slope of the yield curve, and on the demand for ARMs as investments relative to the demand for ARMs by borrowers.

Exhibit 25–3 also quotes the annual percentage rate (APR), which is higher than the original contract rate and accounts for all costs including interest, such as any origination fees, discount points and private mortgage insurance that may be part of the loan agreement.

HISTORY OF SECONDARY MORTGAGE MARKET

The Federal National Mortgage Association (Fannie Mae) was established in the 1930s in order to purchase and sell FHA-insured loans and eventually VA loans (1948). Prior to the 1950s, the secondary mortgage market consisted exclusively of the sale of whole loans to investors. Mortgage banks and thrift institutions would originate mortgages and sell them to life insurance companies, pension funds, and other financial institutions. The explicit government guarantee of FHA and VA loans facilitated the active trading of whole loans in the secondary mortgage market.

For much of its life, Fannie Mae operated as a national savings and loan in the sense that it gathered funds by issuing its own debt and buying mortgages that were held in its portfolio until 1968 when it was divided into two entities. A new GSE, which retained the same name, Fannie Mae, was created as a shareholder-owned government-sponsored enterprise (GSE). Ginnie Mae (GNMA) was also created to continue to provide a secondary market for government-insured loans. In 1970, Ginnie Mae issued its first mortgage pass-through security, in which the timely payment of interest and principal were guaranteed to investors as well as the full repayment of principal even in the event of a default. At this point, Fannie Mae shifted its focus from government-guaranteed mortgages to non-FHA/VA mortgages, known today as conventional mortgages.

In 1970, the U.S. government chartered the Federal Home Loan Mortgage Corporation or Freddie Mac in response to credit shortages for single-family mortgages, which accounted for the largest share of the residential mortgage originations. The government’s objective was to provide a secondary market for thrifts and other originators of conventional mortgages. Freddie Mac was also authorized to purchase FHA/VA loans. At the same time, Fannie Mae was given the authority to purchase conventional mortgage loans, which would allow both GSEs to compete for all residential mortgages below the conforming loan limit.

MBS issuance remained relatively steady in dollar terms for Ginnie Mae from the mid 1980s through 2000; meanwhile, the volume of Fannie Mae and Freddie Mac issuance began to grow significantly in the late 1990s and early 2000s, resulting in a diminishing GNMA share until the financial crisis of 2007–2009. At that point, the non-agency securitization market (also called the “private-label securities,” or PLS, market), which had been supplanting the GNMA share, ceased to function, while the FHA increased loan volume significantly to help ameliorate the shortage in credit. This led to an increase in GNMA issuance for 2008–2010.

A distinguishing characteristic of agency MBS is the implicit or explicit guarantee they carry. For example, Ginnie Mae guarantees the timely payment of principal and interest on all Ginnie Mae pass-through securities backed by the full faith credit of the United States government. Therefore, in the event that a borrower defaults on a mortgage in the underlying collateral pool, investors will continue to receive the timely payment of principal and interest.

In the case of Fannie Mae and Freddie Mac, MBS investors are also guaranteed the timely payment of principal and interest whether or not payments are made by mortgagors. However, the Fannie Mae and Freddie Mac guarantees constitute corporate guarantees and are not backed by the full faith and credit of the U.S. Government. Yet despite the lack of a formal government guarantee, Fannie Mae and Freddie Mac MBS have traditionally been viewed by the financial markets as having credit status better than triple-A bonds (often referred to as the implicit guarantee). The source of this perception is generally attributed to the close ties between the government and the GSEs. The close ties include the history of the GSEs as government agencies, the unique regulatory structure of the GSEs, a largely symbolic line of credit to the Treasury and the presence of presidential-appointed board members for the GSEs.

The guarantees on principal and interest repayment, together with the large issuance this has allowed and a special forward delivery/repo mechanism that we discuss later, have made GNMA and GSE securities by far the most liquid asset-backed security, with significantly greater liquidity than corporate bonds, and at times even greater liquidity than certain U.S. Treasury issues.

The notion of an implicit government guarantee for the GSEs was severely tested during the financial crisis of 2007–2009. Fannie and Freddie began to suffer significant delinquencies in their guarantee portfolios, which were generally expected to result in losses greater than the guarantee premiums they had charged to bear this risk as well as their capital buffers; additionally, they had built up sizable positions in the AAA-rated tranches of nonagency securities, which began to suffer significant price declines as a liquidity crisis hit this market. As many investors began to question their continued solvency, the GSEs ability to rollover their short-term debt came into question, and in an emergency action, the secretary of the Treasury directed their regulator to take them into conservatorship, effectively making the guarantee on agency MBS and senior debt securities explicit. As of writing in early 2011, the ultimate fate of the GSEs is being re-thought, especially the government guarantee. Nevertheless, we expect agency MBS to be a major fixed income asset class for the foreseeable future.

AGENCY POOL PROGRAMS

The GSEs rely on a network of lenders that originate loans in order to create mortgage pools backing their mortgage-backed securities. Qualified lenders typically sell mortgage loans to the GSEs through one of two channels: flow and bulk; lenders are qualified if their origination and servicing practices have passed the GSEs requirements. The flow channel allows volume to fluctuate based on the lenders ability to originate, but with guarantee fees set in advance (prior to 2006, these were set annually, but subsequently have been set quarterly). The guarantee fees are set according to loan-level variables that determine credit quality. The flow channel can be viewed as the most liquid forward market. The bulk channel involves spot sales of loan packages with guarantee premiums negotiated on a per-package basis.

While loans in a pool may share broad similarities, differences do exist. For example, not all borrowers may be paying the same coupon. The loans may not be the same age when they were bundled. In recognition of these differences, aggregate pool characteristics or indicatives are calculated to provide investors with the starting point of their analysis. The standard terminology used to describe a typical agency pass-through can be found in Exhibit 25–4, which contains information about a specific Freddie Mac pool.

Gross weighted-average coupon (gross WAC) refers to the actual dollar weighted average coupon of the mortgages in the underlying loan pool, while the net WAC represents the gross WAC less any servicing fee paid for processing the loan. The WAC of the Freddie Mac pool in Exhibit 25–4 is 8.0%. The proceeds from a mortgage pass-through can be stated on a gross or net basis. The net WAC for this pool is equal to the net coupon of 7.5%. In this particular example, the servicing fee is 50 basis points, which is standard for Freddie Mac pass-throughs. Servicing refers to a third party who processes the homeowner’s monthly mortgage payment, makes payments to local governments for property taxes, pays the property insurance and mortgage insurance premiums if necessary. The servicer also ensures that timely payments are made by the borrower every month and deals with the borrower in the event of delinquency.

EXHIBIT 25–4
Freddie Mac Pool C00875

image

It is important to note that the dispersion of mortgage coupons within the pool is fairly small with limits that depend on the specific agency pool program, which is discussed in further detail in the following. The WAC is important because it not only tells investors about the interest rates of the underlying mortgages, but also reveals the sensitivity of the loan pool to prepayments. When current mortgage rates available to borrowers fall below 1.5% or 150 basis points of the underlying WAC of the loan pool, we would anticipate that the pass-through would exhibit greater prepayment speeds. However, if current mortgage rates were to rise above the WAC of the pool, we would not expect prepayments rates to rise as long as this relationship remained between the WAC and current mortgage rates.

Similar to the WAC, the weighted average maturity (WAM) represents an average maturity weighted by the loan balances of the pool. The WAM is important because it gives investors an idea of how many payments are remaining before the principal of the pool is retired. In this particular example, the WAM of this Freddie Mac pool is 296 months. The weighted average loan age (WALA) is just the converse of WAM and represents the average age of the underlying loans in the pool weighted by their balances. In this example the WALA is 54 months.

The basic characteristics of loans within a mortgage pool have generally been standardized depending on the specific agency and program. The specific pooling requirements for each of the agency programs are summarized in Exhibit 25–5 with outstanding volume through the end of 2010.

EXHIBIT 25–5
Characteristics of Agency Pool Programs

image

image

Each MBS pool carries a pass-through rate or coupon, which is the interest rate passed onto the investor, usually on the 25th day after the end of the accrual period. The pass-through rate is lower than the interest rate on the underlying mortgages in the pool. The interest differential covers the guarantee fee paid to Fannie Mae and the fee paid to the servicing institution.

Ginnie Mae offers several programs: Ginnie Mae I, Ginnie Mae II, and Ginnie Mae Platinum. Each of these programs offers full and timely payment of principal and interest, backed by the full-faith-and-credit guarantee of the U.S. government.

The Ginnie Mae I program issues pass-through MBS where registered holders receive separate principal and interest payments on their certificates. The MBS are based on single-issuer pool. Each pool must consist of a minimum of three loans totaling a minimum dollar amount of $1 million. The underlying mortgages have roughly the same maturities and interest rates. Only fixed-rate mortgages can be submitted under this program. The single-family pools have a 50 basis point guaranty and servicing fee. Payments are made to the holders on the 15th day of each month.

The Ginnie Mae II program, which was introduced in 1983, allows registered holders to receive an aggregate principal and interest payment from a central paying agent on all of their Ginnie Mae II MBS. The Ginnie Mae II program provides greater flexibility to issuers because it accommodates both multiple-issuer pools as well as single-issuer pools. Smaller issuers who do not meet the minimum dollar pool requirements of the Ginnie Mae I program are able to participate with pool sizes as small as $250,000. The Ginnie Mae II program also allows both fixed- and adjustable-rate mortgages. Coupon rates on underlying mortgages can vary between 50 and 150 basis points above the interest rate on the pool. Ginnie Mae II MBS pay on the 20th day of each month.

Ginnie Mae Platinum securities are issued under the Ginnie Mae Multiclass Securities Program. A Ginnie Mae Platinum security is created by combining Ginnie Mae MBS pools with uniform coupons and original terms to maturity into a single certificate. An interesting feature of the Platinum program is that investors owning several MBS with relatively small remaining balances have the ability to aggregate the MBS. This provides critical liquidity to investors. Platinum pools can also be used by investors as collateral for repurchase agreements and structured financial products.

Fannie Mae MBS pays interest on the 25th day of each month (after the accrual period). The pass-through rate is lower than the interest rate on the underlying mortgages; this interest differential covers the guaranty fee paid to Fannie Mae (as well as the servicing fee paid to the servicer). When the underlying loans are pooled together, Fannie Mae permits the interest rates on the loans to fall within a 250 basis point range.

Freddie Mac also offers a pass-through program that offers full and timely payment of interest and ultimate payment of principal guaranteed by Freddie Mac and not by the full faith and credit guarantee of the U.S. government. The Freddie Gold program, on the other hand, offers full and timely payment of interest and scheduled principal guaranteed by Freddie Mac. The Gold program is very competitive with the Fannie Mae MBS program; hence, it is Freddie Mac’s most popular MBS program. The Gold program pays interest on the 15th day of every month. Fannie Mae offers a Mega program and Freddie Mac offers a Giant program that are similar in nature to Ginnie Mae Platinum program as described.

TRADING CHARACTERISTICS

The timing and amount of cash that exchanges hands in a trade involving agency MBS are critical in determining the price and yield of the security. Therefore understanding market trading conventions established for agency MBS is critical in analyzing the relative value and risk of these securities. The key issues are discussed in the following.

Settlement and TBA

Typically, pass-through MBS trade on a forward basis, in which settlement occurs once per month. Each type of mortgage is assigned a particular day during the month for trade settlement. During any particular calendar month, the active month for which most trades will settle will be the next monthly settlement for that security. For example, during the month of June the active trading month will be for July settlement. By the middle of July, traders will generally shift the active settlement month to August. Secondary collateralized mortgage obligations (CMOs) the subject of Chapter 26, trade on a corporate (T+3) settlement basis unless specified otherwise at trade initiation.

The forward market facilitates the origination of mortgage loans, as originating firms can sell MBS forward, prior to creating mortgage pools. In this way, the originators hedge the rates they have “locked” for borrowers. Active trading in the one-month-forward market also stems from the importance of the CMO market. Among most Wall Street dealers, frequently the biggest trading counter party of the MBS pass-through trader is the dealer’s primary CMO desk. The pass-through trader will be responsible for purchasing the CMO collateral needed for any deal. As most CMO deals settle one (or more) calendar month from the pricing date, trading for collateral is most active in the one-month-forward market. This does not preclude other settlement possibilities for investors. Up until two days before the settlement within any particular month, it is still possible for investors to purchase bonds for current-month settlement. Dealers will still make markets for current-month settlement, but not always with the same liquidity as the most actively traded month.

However, there are times when attractive opportunities arise for investors as current settlement approaches. In cases in which a dealer still needs collateral to settle a CMO, he or she may have an aggressive bid for current settlement collateral or be willing to create an attractive spread in the mortgage “roll” market. Investors who roll their mortgage pass-throughs enter into a repurchase agreement in which they sell their MBS today for repurchase at a later date. Exhibit 25–6 shows TBA bid and asks prices for a few FNMA 30s starting with settlement in February 2011 given the trading environment as of January 26, 2011. For example, assuming execution at the mid-point of the bid and ask, rollers of FNMA 4.5s would sell at 101-28.5/32 for Feb settlement and repurchase at a lower price of 101-17.5/32 for March settlement. The difference is referred to as the drop and can be thought of as a partial coupon payment received by the lender. The column labeled Drop gives more precise bid/ask levels for the drop. The units are in 32nds, so for example, the Feb/Mar 4.5 coupon drop bid is (10 5/8)/32 which in decimal form is approximately 0.33. Each drop, assuming a particular one month prepayment speed on the underlying TBA, corresponds to an implied financing rate.

EXHIBIT 25–6
Sample of TBA Rolls (Sample Dollar Rolls & Financing Rates)

image

These financing rates are shown in the final column. The greater Wall Street’s demand for mortgage pass-throughs settling in the nearby month, the greater is the drop. The demand for current production pass-throughs is driven by the need for collateral for CMO production or other investor demand for MBS. The choice of rolling bonds will depend on whether investors can reinvest the proceeds of the sale of their mortgage pass-throughs and earn a return greater than the yield of the bonds. The greater the drop, the greater the implied financing rate of the bonds is.

Although the roll is very similar to a reverse repo transaction in other government bond markets, there are some differences worth noting. The investor who sells the bond is not likely to get the same pool back, but one with substantially similar characteristics. In general, the buyer will deliver back the cheapest bond they can find that still matches the coupon and TBA guidelines, so when running a roll analysis, it may be best to assume the most adverse pool will be delivered back to you. Determining the cheapest to deliver bond can be substantially more difficult to anticipate than in the government securities market due to the complexities of prepayments and its drivers.

Settlement Cash Flows and Security Delivery

At the time of the trade, the two counter parties agree to the settlement date, price, and quantity of securities. At the time of settlement, the purchaser will pay the price times the quantity of securities, plus any accrued interest. The interest-accrual period will include the number of days from the first of the month until the settlement date.

Most trades are on a to-be-announced (TBA) basis. Mortgage lenders are allowed by the agencies to sell mortgages forward by securitizing the mortgages for purchase in the secondary market. In order to allow lenders to hedge (or fund) their origination pipelines, settlement dates are set between one and nine months from the date on which the transaction is negotiated. This permits lenders to lock in a price for the mortgages for which they are in the process of originating. An interesting feature of the TBA market is that the purchaser does not know what pools will be delivered until just prior to settlement. The number of pools that will be delivered (and the precise characteristics of the pools) are unknown. Generally, TBAs are analyzed using the average characteristics of the given mortgage program.

Trades can also occur on a specified pool basis. These trades may reflect special inventory that a dealer holds or that a client needs. Trades on specific pools usually occur at prices above the current TBA quotes due to some valuable properties that the specified collateral has in comparison to TBA, or cheapest-to-deliver, collateral. In addition to specific pools, buyers and sellers may negotiate other types of characteristics such as year of origination or number of pools to be delivered.

Specified pool trading occurs for seasoned or WAM bonds, lower loan balance pools (LLB), lower FICO or Alt-A pools, etc. Due to their prepayment characteristics relative to the average pools (generally less sensitivity to interest rate volatility) WAM bonds tend to have greater value than the generic pool and, therefore, sell at higher prices than the TBA price. Wall Street traders also use specified pool trading to obtain pools for structured transactions or reduce back office costs by restricting the number of pools to be delivered.

While specified pool trading has existed throughout the life of the agency MBS market, the proportion grew tremendously between 1995 and 2003, which led to considerable fragmentation of the agency MBS market. The fragmentation is a result of the large range of loan sizes and dispersion of credit quality of borrowers within Fannie Mae and Freddie Mac conventional 30-year loan pools, combined with the disclosure of more borrower variables in more detail over time. The differences in agency pool composition can result in substantial differences in prepayment sensitivity and consequently, investment decisions. During periods of high demand, pools that were purchased on a pool specific basis are often delivered into TBA transactions, thus providing a contingent source of liquidity for the TBA market.

In the past, Fannie and Freddie provided information about loan coupons, remaining term, weighted average loan age, weighted average loan term, original WAM, original loan size, and issuer for each pool. Starting in 2003, they began to disclose additional pool characteristics, which began to impact mortgage trading and prepayment modeling. The additional disclosures include loan purpose, original loan-to-value (LTV) ratio, standardized credit scores of borrowers (FICO score), servicer, occupancy status, geographical distribution (at the state level), and property type. The availability of these additional pool characteristics led to an increase in the choice of specified pools, and has resulted in a greater ability to select cheapest-to-deliver pools that contain more prepayment sensitivity.

Delay of Cash Flows

Nearly all borrowers make their mortgage payments in arrears on a monthly basis. Likewise, investors receive their cash just once a month. The time between the expected cash payment from the borrower and the ultimate cash flow received by the investor is called the delay. The effect of this delay must be treated by the yield calculations performed on MBS because it represents a true loss of economic opportunity. The delay varies slightly among the agencies and GSEs (see Exhibit 25–5).

While the delay factor is meant to cover many of the exigencies that occur when borrowers are late with their payments and the mechanical complications of processing the cash flows, it also provides an important source of income to the financial intermediaries. Both Fannie Mae and Freddie Mac derive significant income from the float earned between the time they collect cash flows and disburse them to investors.

Accrued Interest

The MBS begins to accrue interest on the first calendar day of the month. This corresponds to the same accrual period for the borrower. At the time of settlement, the investor must pay the previous holder the interest through the settlement date. After settlement, the investor is entitled to the entire month’s interest.

MBS accrue interest on a 30/360-day basis. That is, accrued interest calculations assume that each month has thirty days and that each year has 360 days. Practically speaking, this means that each month the investor receives 1/12 of the annual coupon. As for calculating accrued interest, the investor receives 1/30 of the monthly interest payment for each day up until settlement. No additional interest is paid for settlement on the 31st of the month. Typically most settlement occurs in the middle of the month, so the extra day is not an issue.

Delivery Standards: Variance and Pools Per Million

In the nuts and bolts of trading MBS, some accommodations are made to smooth the settlement process. Many MBS pools are not originated in round dollars. However, trades between dealers and institutional investors usually take place in even lot sizes of one million dollars or more. To accommodate the anomalies of pool size, the seller has some flexibility regarding the number of pools delivered and principal amount settled.

In settling a trade, the seller can modify the amount of principal delivered. This deviation from the original trade amount is called variance. The variance permitted on TBA trades of Fannie Mae, Freddie Mac, and Ginnie Mae securities is plus or minus 0.01% of the dollar amount of the transaction agreed upon by the parties. In the case of specified pool trades, no allowance for variance is permitted. If the variance is not within the 0.01% limit, the pools are not considered good delivery. In a declining market, sellers of MBS will probably deliver as much as possible in a trade, taking full advantage of the upper end of the variance limit. The opposite will occur in a rising market.

In order to keep someone from delivering a large number of low principal dollar pools, the Securities Industry and Financial Markets Association (SIFMA) limits the number of pools delivered in a trade. Restricting the number of pools to a small number has some operational benefits for the MBS purchaser. Tracking the monthly principal and interest payments can be labor-intensive and costly. Left unchecked, sellers would attempt to unload all of their small pools on someone else. Over time, the delivery requirements have been modified somewhat, allowing for more pools per trade for higher coupon MBS. The increased number of pools reflects the high pay-downs and the resulting low balances. The standard requirements for delivery and settlement of MBS are published in the SIFMA Uniform Practices manual.1

PREPAYMENT AND CASH-FLOW BEHAVIOR

Prepayments are the primary distinguishing feature of agency mortgage-backed securities. Without prepayments, mortgages would be extremely easy to analyze. On the other hand, if MBS were so simple, the number and scope of investment opportunities would be severely limited. Prepayments are the double-edged sword of the MBS market. They create opportunity, but they also create risk.

The timing and amount of cash flows received by pass-through MBS are greatly affected by the prepayment behavior of underlying mortgages within the MBS pool. Borrowers generally have the right to prepay their loans at any time without penalty (with the exception of some subprime loans). Borrowers may pay off their loans in full or in part. For a full prepayment, the borrower pays off the remaining outstanding balance of the loan, usually using the proceeds of another loan at a lower rate. Partial prepayments, often referred to as curtailments, reduce the balance of the loan, but do not alter the scheduled monthly payments. For individual loans, a partial prepayment shortens the life of the loan. However, for a pool of loans, a full prepayment of one or several loans does not have the same effect as a partial prepayment of an individual loan. Full prepayments of loans within a pool serve to reduce the outstanding balance, but do not reduce the final maturity of the underlying loan pool.

PREPAYMENT CONVENTIONS

Prepayments are defined as the difference between the actual balance of the MBS pool in any month and the balance expected due to normal amortization. The market has developed several approaches to describing the prepayments for a pool of loans. Since MBS pools vary in size, prepayments measured in dollar terms would not make meaningful comparisons. Exhibit 25–7 summarizes three prepayment conventions, which are all expressed as a percentage or rate.

EXHIBIT 25–7
Prepayment Conventions

image

In order to develop a familiarity with the conventions, we will work through some numerical examples.2 Using information about the scheduled balance, actual balance and age presented in Exhibit 25–8, we can calculate the prepayment rates in SMM, CPR, and PSA formats.

EXHIBIT 25–8
Sample Prepayment Rate Calculations

image

SOURCES OF PREPAYMENTS

Prepayments arise when borrowers move, refinance, or default on their mortgage loans. Prepayments that result from changing residences constitute a base prepayment rate, which is referred to as turnover. Employment related relocation, marriage, divorce, children, death of a spouse, and other life cycle changes are factors that directly affect housing turnover. Most conventional mortgages (non-FHA/VA loans) contain a due-on-sale clause, which stipulates that the mortgage must be paid in full when the house is sold. In contrast, FHA/VA loans are often assumable, which means they can be transferred to the new homeowner as long as the new borrower meets minimum credit requirements.

Mortgage refinancing represents the largest and most variable source of prepayments. There are several categories of refinancing that can occur. For example, there are high quality borrowers who want to take advantage of lower cost mortgages in a falling interest rate environment. There are also borrowers who refinance in order to borrow more money than the existing loan balance on their property, provided that there is sufficient equity in the home. This type of refinancing activity is called a cash-out refinance. Borrowers with previously tarnished credit histories are able to refinance at more favorable rates because of improvements in their credit ratings. This type of refinancing is referred to as credit curing.

Defaults are not technically prepayments, but they have the same effect as prepayments in that the principal balance of the defaulted loan is returned to the investor in the case of agency MBS. Because of the guarantee provided by the GSEs, the investor is protected from the credit risk of individual borrowers that compose the pool. Defaults of agency MBS typically represent only a small fraction of monthly prepayments because of the high credit quality of the underlying mortgages, and therefore, generally can be forecast as part of prepayments. Notable exceptions were agency pools composed of Alt-A low-doc loans originated during the period 2006–2008. In such cases, it is critical to forecast defaults separately from prepayments because of the lower quality of the collateral.

The most important factors that impact prepayment behavior include interest rates, aging, loan size, burnout, and seasonality. The primary factor influencing prepayment rates is the level of interest rates, which reflects a borrower’s opportunity to refinance. The refinancing incentive can be measured by the difference between the interest rate of the borrower’s loan versus the interest rate currently available in the market for new loans. For example, in the Andrew Davidson & Co. prepayment model, current coupon yield (the yield on MBS trading near par) is used as a proxy for the rates currently available to borrowers. Loans in which the difference between the loan coupon and the current coupon yield are greater than zero have a greater incentive to refinance. Exhibit 25–9 demonstrates the effect. As the refinance incentive moves from negative to positive territory, prepayment speeds increase as shown by CPR along the vertical axis. The relatively level prepayment speeds for refinancing incentive less than zero reflect the base turnover rate in the housing market. Such loans, in which the coupon is less than the current coupon yield, are called discount loans because they have prices less than par.

EXHIBIT 25–9
Interest-Rate Effect on Prepayment: GNSF

image

The response of prepayments to interest-rates is decidedly nonlinear. (Many early prepayment functions used an arctangent function to fit the S-shape we see.) Most academic papers that have attempted to value MBS in simple ways (without using Monte Carlo simulation, which is discussed in the section on valuation) assume a linear response, is makes these academic approaches to date impractical for use by practitioners.

Because refinancing into a new mortgage comes with fixed costs (such as an appraisal, origination fee, etc.), the size of a loan is an additional factor affecting the refinancing incentive perceived by the borrower. All else equal, on a percentage basis, these fixed costs are lower the larger the size of the loan, leading to steeper refinance incentive curves than shown above for Fannie and Freddie loans, which tend to be larger than FHA loans. The refinance incentive curve for prime jumbo loans is even steeper.

Aging is the second most important factor influencing prepayment behavior. Aging reflects the fact that new loans typically exhibit slower prepayment speeds compared with “seasoned” or older loans. The PSA benchmark aging curve was introduced in the mid 1980s to account for the aging pattern typically observed for agency MBS. The base PSA curve or 100% PSA is depicted in Exhibit 25–10 and starts at 0 CPR and rises linearly to 6 CPR by month 30. The PSA curve roughly depicts the prepayment pattern for discount loans but not for premium loans. Because of the variation in aging patterns for different loan types (in particular the substantial variation by refinance incentive), the PSA curve should be considered an outdated convention. We have included it here for completeness because a number of popular analytical systems continue to include it as an alternative to CPR and SMM.

EXHIBIT 25–10
PSA Prepayment Convention

image

Exhibit 25–11 shows the aging pattern of Ginnie Mae loans by coupon type. Premium coupon loans are defined as loans whose coupons are greater than the prevailing market rate and therefore have a positive refinancing incentive, while discount coupon loans have coupons less than the current market rates. The exhibit illustrates that newer loans prepay more slowly, regardless of whether they are premium or discount loans. Moreover, the exhibit also shows that not all loans follow the same prepayment trajectory. The premium loans experience higher prepayment speeds and peak sooner compared with the discount loan counterparts in the exhibit. In addition, the prepayment rates of discount loans tend to rise more slowly and peak farther out on the aging curve.

EXHIBIT 25–11
Prepayment Rate of Discount versus Premium Loans

image

Burnout is another important factor determining the rate of prepayments and probably the most difficult to measure. The concept of burnout is based on empirical observation that refinancing activity within a loan pool declines over time regardless of whether interest rates continue to decline further. The burnout effect can best be understood by viewing an MBS pool as a collection of borrowers, each with different propensities to refinance. As rates fall, the borrowers with greatest propensity do so. The borrowers remaining in the pool face higher refinancing costs, economic and noneconomic. These borrowers tend to repay at a slower rate compared to the first group of borrowers even when later faced with the same interest rate differential. Because the composition of the pool depends on the path that interest rates and refinancing have taken since the pool was originated, burnout is called a path-dependent variable. In addition to the nonlinear response to refinance incentive discussed earlier, this path-dependence presents additional challenges in valuing MBS compared to other asset classes.

Exhibit 25–12 illustrates both the concept of burnout and the difficulty of measuring it using data from the aggregate performance of an FNMA 30-year fixed pool from 2002 with performance from 2002 through the end of 2010.

EXHIBIT 25–12
Comparison of Prepayment Speeds, FNMA 30-year 2002 Vintage

image

The bars show the total balance of that vintage over time, the dashed lines show our model forecasts (given an accurate interest rate forecast), whereas the solid lines show actual aggregate prepayment speeds. As we can see, the majority of the balance of the 2002 vintage refinanced during the strong incentives of 2003 and 2004, and the model was largely accurate during that period. Later, with a very changed pool composition, despite rates dropping much lower during the financial crisis (to all-time lows in early 2009), the model forecast much lower prepayment peaks, topping around 50 CPR. Nevertheless, this vintage in aggregate behaved much more burnt out, primarily due to the extremely difficult housing market and credit crunch. (This is discussed in more detail in the paragraph on housing and the general economy influencing prepayments.)

Seasonality is another major determinant of prepayment rates. Seasonality reflects the close interaction of prepayments with housing market activity. Historically, prepayments tend to be faster during the summer and slower during the winter months, reflecting increased home turnover during the summer months. Seasonality is more pronounced for discount loans compared with premium loans. Premium loans do not exhibit a large seasonal component because refinancing activity tends to eclipse the seasonality effect. The commonly observed features of seasonality are as follows: Prepayments tend to be highest in May and July with an uptick in December, perhaps reflecting year-end tax strategies and moving during the holidays, especially summer vacation.

House prices and the general economy also influence prepayments. Weak economic activity, declining housing prices and unemployment all tend to depress prepayments. These factors are sometimes regional in nature and can lead to large differences in prepayments across the United States. During the 1990s, home price appreciation strongly affected prepayments because rising home prices resulted in greater homeowner equity, thus increasing financial opportunities available to borrowers for cash-out refinancing. Later, during 2006–2010, as home prices declined nationally, prepayments were dramatically depressed, even at what on paper looked like comparable refinancing incentives. Many borrowers had insufficient equity to allow them to refinance; simultaneously, banks, and the GSEs became much more cautious in their underwriting, and raised the premiums charged for taking credit risk. This resulted in a long period of depressed prepayments relative to the late 1990s and early 2000s, when strong house price appreciation resulted in faster prepayments at given refinance incentives than people had come to believe based on prior experience. This period has led some modelers to include an explicit credit-spread in their measures of refinance incentive.

The shape of the yield curve can also impact prepayments. With the development of a variety of mortgage products such as ARMs, hybrid ARMs, and balloon loans, borrowers can choose products whose prices are derived from different parts of the yield curve. When the yield curve is steep, borrowers may refinance into shorter maturity loans in order to reduce their borrowing costs. When the yield curve is flat, it may be advantageous for borrowers to lock in rates at the long end of the yield curve particularly if they have a long time horizon.

PREPAYMENT MODELS

Evaluation of the investment characteristics of MBS requires estimates of prepayment rates. These estimates can take various forms, from a single assumption based on experience to complex models that rely on loan level details.

These forecasts, regardless of their source, are used to understand the performance characteristics of MBS and determine appropriate valuation of different investments. Prepayment forecasters face a fundamental problem. They seek to estimate future events in a changing world. For example, new loan types are constantly being created and the loan origination process is continually evolving. In addition, over time, our understanding of the primary variables affecting prepayments and their interactions has continued to evolve. Still, the primary guide to future prepayments is past prepayments. Thus modelers develop models which seek to explain prior prepayments well based on a parsimonious set of variables while retaining some flexibility to allow for changes in understanding that occurs over time. Since the economic and social environment is constantly changing, and prepayments are affected by a host of factors, it is unlikely that any historically based analysis will completely reflect future prepayments. Investing based on prepayment models is a little like driving, while looking through the rear view mirror. It may be hard to stay on the road, but it’s better than driving with your eyes closed, especially on stretches where the road does not turn too dramatically.

Good models are ones that are robust and parsimonious. Robust models provide good forecasts under a variety of conditions. That is, they do not need to be adjusted continually in order to reflect changing environments. Models that require constant adjustments probably will not provide accurate forecasts of future prepayments. Parsimonious means that the models are as simple as possible. Parsimonious models capture the major variables that affect prepayments using the fewest number of parameters. They have the advantage that they do not “over fit” the data compared with models with too many factors. The use of complex models with too many parameters may result in an excellent fit to historical data, but may not provide accurate projections. The added variables may reflect a spurious one-time correlation rather than real long-term relationships. Parsimonious models are also easier to incorporate into valuation tools and explain to a wide audience.

VALUATION

A common measure of value quoted among investors is the spread between the MBS and a benchmark U.S. Treasury security. While spread measures are inadequate because they provide only a static view, they are a useful starting point for assessing relative value. Exhibit 25–13 presents various spread measures for Fannie Mae 30-year with coupons ranging from 4.5% to 6.5%.

EXHIBIT 25–13
Spread Analysis for Fannie Mae 30-Year, January 14, 2011

image

The first spread measure presented in Exhibit 25–13 is the static spread, which is labeled Spread/Benchmark. It is computed by subtracting the yield of a specific U.S. Treasury security from the yield of the MBS. For newly issued current coupon MBS, the benchmark Treasury is often the on-the-run 10-year Treasury note. Other MBS are compared to U.S. Treasuries with similar average life. The first number represents the spread in basis points and the second number represents the benchmark Treasury. For example, the spread of FNMA 5.0 is 197 basis points over the 3-year Treasury.

A more reliable measure of value than static yield is an interpolated weighted average life spread, which is also shown in the same exhibit as “Spread to WAL Treasury.” This measure uses a Treasury benchmark that more closely reflects the principal repayments of the MBS. For example, rather than compare the FNMA 4.5 with a 5-year Treasury, we would compare this MBS pass-through with a Treasury with the same WAL of 3.7 years. In order to do this, we interpolate the yield of a Treasury with an average life of 3.7 years. Notice that the interpolated weighted average life spread for the FNMA 4.5 is greater than the static spread. The difference can be attributed to the divergence between the average life of the MBS and benchmark Treasury, which makes the WAL spread more reliable.

The spread to the zero curve or “zero spread” is also presented in Exhibit 25–13, which is a great improvement on static yield and interpolated WAL spread because it incorporates the shape of the yield curve and timing of cash flows. The cash flows of the MBS are discounted by the zero coupon rate plus a constant spread that equates the net present value of cash flows to the current market price. In Exhibit 25–13, the zero spreads are lower compared with the WAL spreads. The reason for this is that lower yield spreads earned on longer dated cash flows have a greater effect on value than the higher spreads earned on earlier cash flows. The flatter the yield curve, the less divergence will be observed between WAL and zero spread measures.

All of the spread measures we have mentioned so far suffer from the deficiency that they are all computed based on a single interest rate scenario. In reality, we are never able to forecast the path that interest rates will take. Therefore, a measure of value that recognizes this uncertainty is needed.

Option-adjusted spread (OAS) is such a measure of value because it takes into consideration the changing nature of underlying cash flows under a multitude of interest-rate scenarios. OAS represents the expected spread over the short-term risk-free rate over a large number of possible interest rate scenarios after accounting for the prepayment option. OAS captures the additional yield an investor is expected to earn from the prepayment option component of MBS cash flows. The “expected yield” from an MBS can be thought of as the risk-free return (over the life of the mortgage suitably time weighted) plus option cost plus OAS.3 In general, OAS calculations from price, or price calculations from OAS utilize Monte Carlo Simulation. This is in contrast to the valuation of other, simpler instruments which might be done in closed form (using a “formula”) or using backward induction (a faster numerical approach than Monte Carlo). The nonlinearity of prepayments as well as path-dependence prevents the use of the first approach, whereas path dependence often prevents the use of the second.

Additional measures such as effective duration and effective convexity provide insight into the market risk of the security. Effective duration measures the change in price for a 100 basis point change in interest rates, holding OAS constant. Convexity measures the change in duration for a 100 basis point change in rates. Negative convexity is a general feature of MBS (or other bonds with embedded optionality), reflecting that the amount of upside price appreciation is limited relative to the amount of downside price risk, as a function of interest rates. The reason for the use of the term effective is to distinguish these measures from static (single-path based) measures, which happen to coincide for simpler securities (such as noncallable Treasuries) but would give inaccurate results for MBS.

Exhibit 25–14 presents OAS results for various coupon Fannie Mae 30-year pass-throughs. The calculations for the starred line were generated using the Andrew Davidson & Co., Inc. prepayment and OAS models. The OAS values on Fannie Mae pass-throughs on December 17, 2010 range between 5 and 35 basis points using our model. The dashed and triangle lines show the median and mean OAS values, respectively, using a sample of results from five major dealers that publish weekly OAS results. Discrepancies on the order of 20 bps between different models regularly occur, highlighting the difference primarily caused by the use of different prepayment models.

EXHIBIT 25–14
FN30-Comparison of AD&Co and Dealer OAS Pricing Date December 17, 2010

image

We show a similar comparison of option-adjusted durations (OAD) from the same valuation date in Exhibit 25–15. Similar to the OAS results, variation between different firms’ results is primarily attributable to prepayment model differences (with an additional contribution from interest rate simulation models). The exhibit shows that the duration of discount coupons (4 and 4.5s) grows quite large, and at increasing slope, whereas the rate at which duration shrinks in the higher coupons (5.5 through 6.5s) diminishes, with a flattening of the duration curve.

EXHIBIT 25–15
FN30-Comparison of AD&Co and Dealer OAD Pricing Date December 17, 2010

image

This feature is caused by the negative convexity of MBS. This means that gains in decreasing interest rate environments will be limited compared to fixed income investments with positive duration, whereas losses in increasing interest rate environments will be greater. The option cost yield premium paid by MBS can be viewed as the compensation paid for this feature of MBS whereas the OAS premium itself can be viewed as the compensation paid for the prepayment model uncertainty.

KEY POINTS

• Mortgage-backed securities consist of pools of mortgages with cash flows to investors dependent on underlying mortgage cash flows as well as potential cash flow structuring rules (CMOs). The simplest MBS are pass-through securities.

• The underlying types of mortgages can be a variety of fixed-rate mortgages, ARMs, and hybrid ARMs, with a choice of amortization structures, underlying indices and cap structures. Fixed-rate 30-year amortizing mortgages remain the dominant form of mortgage loan, but hybrid ARMs have also become popular in the last decade.

• Agency MBS are those issued by GNMA, Fannie Mae, and Freddie Mac.

• Prior to the financial crisis of 2007, Fannie and Freddie carried an implicit government guarantee while GNMA has always carried the full faith and credit of the U.S. government. During the financial crisis, Fannie and Freddie entered conservatorship, effectively making the guarantee of agency MBS explicit. Although their ultimate resolution and future form may not be decided until 2012 or later, it is expected that agency MBS will remain a major asset class for many years to come.

• Investors are attracted to agency MBS due to their high credit quality, yield pick-up, and tremendous liquidity.

• The difficulty of modeling prepayments makes the valuation and risk management of MBS challenging, even compared with the analysis of other fixed income instruments with embedded options. In particular, the path-dependence of mortgage pools and CMOs, together with the nonlinearity of prepayment models, makes Monte Carlo simulation the preferred approach for valuation among practitioners.

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

You can't read the all page of ebook, please click here login for view all page.
Reset
34.201.28.181