The following article is reprinted from the January/February, 2002 issue
of On the Edge, the Interactive Data Fixed Income Analytics bimonthly newsletter.
Forecasting Mortgage Rates in Today's Environment
Kamel Bazizi, Vice President of Research & Consulting
Investors rely on Option-Adjusted Spread (OAS) models for the valuation of mortgage-backed securities. One of the most important parameters in an OAS model is the prevailing mortgage rate, which is used in conjunction with a prepayment model to determine the refinancing incentive component of prepayment forecasts for various mortgage products. Prepayment cash flows are projected and used to price mortgage backed securities and determine their risk measures. Since a daily mortgage rate is not readily available, it is commonly estimated based on the market par coupon rate plus the spread between primary and secondary market rates. The par coupon rate is computed from live TBA mortgage prices, while the spread is estimated from historical data. The spread represents the fee charged by mortgage lenders over the secondary market rate. In the current low rate and high volatility environment the spread is very wide by historical standards. For the week ending 10/26/01, it measured 61 bps, 30 bps above 5-year average and 21 bps above the 2001 average. In this article, we first discuss the impact of spread widening on mortgage pass-through durations, then explain the factors driving the spread.
The Fed’s announcement on 10/31/01 to halt issuance of the 30-year Treasury led to a substantial bond market rally. The 30-year FNMA par coupon rate dropped to 5.76%, a new 30-year low. Of course, secondary market rates, like par coupon rates, can be misleading indicators of what a homeowner actually pays. In fact, the mortgage rate computed based on the par rate plus the latest two-month average spread of 55 bps, was 6.31%. On the other hand, a more reliable rate such as the primary rate published weekly by Freddie Mac measured 6.56% on the week ending 11/2/2001. Consequently, a mortgage rate based on the secondary rate would have been underestimated by 25 bps.
Exhibit 1: Duration change for a +25 bps shift in the mortgage rate
Does it really matter whether to use the primary or the secondary rate?
To address this question, we analyze the effect of mortgage rate changes on the Option Adjusted Durations (OAD) of 30-year FNMA TBA securities, using the BondEdge system. Exhibit 1 shows prices for 5.5%-8.5% coupons as of 10/31/01. The following columns show the corresponding durations and lifetime PSA speeds. We then shifted the 30-year mortgage rate by +25 bps and recomputed new durations and lifetime PSA speeds. The duration change is shown in the last column of Exhibit 1 and graphed in Exhibit 2. It is clear that it makes little difference for deep in-the-money coupons like 7.5%-8.5%s. However, for marginally refinanceable and cuspy coupons like 6.0%-6.5%s, 25 bps make a difference that is as large as 0.3 years in duration. For this reason the daily rate must be estimated as accurately as possible given the available data.
Exhibit 2: Change in FNMA 30 durations for +25 bps change in mortgage rate
Freddie Mac publishes a weekly average mortgage rate, also known as the Survey rate. The rate is the prevailing rate in the primary market between lenders and borrowers. However, for daily valuations of MBS securities, a mortgage model requires a daily mortgage rate. For this reason, daily mortgage rates must be estimated, for example, based on the rate in the secondary market. More precisely, a daily mortgage rate is estimated based the market par coupon rate plus the secondary-primary spread, where the mortgage par coupon rate is implied from mortgage prices and the spread is determined from historical data. We first formally define the par coupon rate, and then discuss the spread in more detail.
TBA pass-throughs are the most actively traded mortgage securities. Live prices are available for TBA mortgages for a wide range of coupons and collateral types. The 30-year mortgage rate is commonly determined from 30-year FNMA TBA prices. The par rate is derived by linearly interpolating between the premium and discount coupons closest to par. The par rate is estimated based on the following equation:
where disc. rate and prem. rate are the coupon rates of the discount and premium coupons, and the disc. price and prem. price are the prices of the discount and premium coupons, respectively. The following is an illustration of the par rate calculation. As of 10/31/01, the 30-year FNMA discount coupon closest to par was the 5.5%, which was priced at 98:24, and the premium coupon closest to par was the 6%, priced at 101:04. Substituting the coupon rates and prices into the equation above leads to a par rate of 5.76%.
While the par rate is computed daily from TBA prices, the spread between the primary and secondary rates is estimated from historical data. The spread and the par rate for the 1991-2001 time period are shown in Exhibit 3. It is interesting to note that the spread is generally much higher when rates are low. The spread was as low as 20 bps in high rate environments, but widened to over 50 bps when rates dropped substantially. Specifically, spreads were the widest during 1993, 1998, and 2001, when rates were the lowest. Furthermore, very low rates are often associated with higher volatility levels, as witnessed by the experience of 1998 and the current environment. The general operating procedure for mortgage lenders is to lock rates for 30 to 60-day periods, which allows borrowers time to close on their new home or on a refinance. By definition, it is more expensive for mortgage lenders to hedge these rate locks in higher volatility environments. This further contributes to the widening differential between primary and secondary mortgage rates. As a result, that differential is widest at low rate levels.
Exhibit 3: Spread between the mortgage Survey rate and the par rate
To understand the relationship between the mortgage spread and rate levels, it is important to understand how these rates are commonly set. Assume that servicing costs 25 bps, and that the guarantee fee charged by agencies is 12.5 bps. Also assume that 30-year FNMA 5.5%s trade at 99 (close to the 10/31/01 closing price of 98:24). This means that new mortgages could theoretically be issued at (5.875% + 2 points). Of those 2 points, 1 point compensates for the fact that loans made at par are sold at 99; while 1 point is for the transactions costs of handling a mortgage, including hedging costs for the rate lock prior to closing. Because very few borrowers want to put down 2 points up-front, a mortgage lender will raise the mortgage rate 25 bps for each point (4:1 ratio). So current pricing translates into (6.125% + 1 point), or (6.375% + 0 points). However, mortgage lenders currently charge on these mortgages about 6.625%, or even higher. So why has the spread between the primary and secondary rates widened much more than in the past? In general, competitive pressures force mortgage lenders to waive some of the costs associated with handling a mortgage. In fact, they are rarely able to charge the full point in transaction costs. But in this historically low rate environment, pipelines are filled up and lenders have no reason to lower their rates. Furthermore, the market volatility stands at historically high levels, leading to much higher hedging costs which are passed to borrowers.
To price mortgage-backed securities and establish an effective hedge strategy, investors rely on option-adjusted durations. This analysis shows that forecasting mortgage rates accurately is an important ingredient in obtaining accurate durations. In the absence of a live feed for daily rates, a model must rely on prices of TBA mortgages to estimate mortgage rates. The mortgage rate may be estimated using the par TBA coupon rate plus the spread between primary and secondary mortgage rates, where the spread represents the fee charged by mortgage lenders over the secondary market rate. The current spread is at a historical high for two reasons: due to low rate levels, mortgage pipelines appear to be full and lenders are under no pressure to lower their fees. Furthermore, high market volatility has caused lenders’ hedging costs to increase, and these costs are passed to borrowers. The widening of the spread between the primary and secondary rates, in these extreme rate and volatility conditions, should be taken into account to successfully forecast mortgage rates.
