Valuation Commentary
Insights
into Historical Current Coupon OAS
By Alex Levin in Collaboration with Andrew Davidson & Anne Ching
When investors attempt to assess how attractive the mortgage market in general is, there exists no better measure with which to begin than the OAS for the current-coupon TBA. This OAS is often found and reported versus the swap rates ("Libor OAS" or LOAS), but the measure taken off the agency debt curve is also informative. A systematic, though volatile, difference between expected returns on agency MBS and agency debentures is a point of investor interest. On the other hand, it serves as the source of income for the U.S. agencies that are the major MBS investors, themselves, holding about 30% of the market (Perli and Sack [2003]). Nevertheless, this positive spread is not a free lunch, it compensates for prepayment risk (even after prepay option exercise is modeled) and operational risk although small differences in liquidity and credit quality exist as well. Let us briefly review these sources.
Risk Factors
Prepayment risk. A large portion of prepayment uncertainty is associated with interest rates and, as such, explained by prepayment models that are inherent to modern OAS analytical systems. If prepayments were perfectly explained by a model, then there would exist little reason for "prepayment risk premium": an option model coupled with a "perfect" prepayment formula should be able to deliver the right price for an agency-backed (default-protected) MBS operating with OAS = 0. An MBS would be valued flat to a known benchmark curve - similarly to swaptions, just with a more complex exercise rule.
Savvy market participants realize that a model can tell only part of the prepayment story. A model's inability to predict prepayments exactly causes unexplained deviations of prepayment speeds above or below the model's forecast, often termed "prepayment surprises" or "prepayment errors". We associate the notion of prepayment risk with this uncertainty unexplained by an unbiased prepay model and assume that the OAS compensates for this risk. AD&Co. maintains a two-dimensional view of prepayment risk: the risk that a model overstates the turnover speed (turnover risk) and the risk that it understates the refinancing speed (refinancing risk). Correspondingly, both premium MBS (exposed to refinancing) and discount MBS (exposed to turnover) should be priced at somewhat elevated OAS levels. This two-risk-factor theory explains well the phenomena of IO/PO pricing too, but what about the subject of our interest, the current coupon MBS?
First instinct is to view the current-coupon MBS as a prepay-risk-free instrument, but this is assumption is amiss. If MBS were valued solely off their yield it would be the right move. However, valuation with OAS accounts for both the curve and optionality and results in a more complex risk picture. Decelerating the turnover, we push the MBS cash flow out causing a value-detrimental effect if the curve is steep, almost no effect if it is flat, and even some value gain if it is inverted.
On the other hand, surprisingly fast refinancing will increase the value of prepay option, which also damages the MBS investor position. However, the steep curve environment is now less risky than the flat or inverted one. With faster refinancing, potential losses are partially compensated for by shorter cash flows (discounted at lower rates) and out-of-the-money steep forward paths. In general, investors in current-coupon MBS should certainly demand compensation for the combination of turnover risk and refinancing risk.
Liquidity and credit quality. TBA MBS are generally as or more liquid than debentures, supply of which may be limited from time to time. MBS and debentures are backed by the full faith of government agencies; in addition, MBS are collaterized by the property. While the credit of agency-backed instruments is typically never an issue, a common perception has been that agency MBS carry even higher credit quality than debentures.
Operational and model risk. Managing and hedging MBS portfolios requires more expertise and costs more to bankers and investors than similar operations with portfolios of bullets. The use of rigorous MBS models is necessary and can't be accomplished without comprehensive data collection and analytical processing. Even after the best models are built or licensed, an MBS investor is not immunized against model errors or a sudden change in market conditions or regulation (such as reducing cost of refinancing and boosting the prepay option value).
Supply and demand. Supply of MBS is driven by rates. When rates fall, supply grows and cheapens. As we will see in the next section, this effect explains the LOAS directionality, i.e. LOAS dependence on the level of rates. The directionality gets smaller if we measure OAS off the agency debenture curve. Indeed, agencies will issue more debts to accommodate for a larger MBS supply. This issuance will make debentures cheaper, too.
Historical Analysis
History of the current-coupon OAS. The use of a single model is required to compile historical OAS retrospectively. This is a preferred way of completing the study to relying on numbers reported in the past. As experience shows, no firm freezes its models for years; each update causes jumps in values (or OAS). The following exhibit (Figure 1) is a single-model retrospective view going back to the times of the Russian crisis of '98. We employed the same prepay model and the Hull-White interest rate model operated with a time-dependent volatility function calibrated each day to the ATM swaptions.
Figure 1
We first observe the LOAS numbers as fairly stable. Despite it being a time of crisis, the standard deviation has remained about 10.6 bp LOAS for the conventionals and 8.6 bp for the governments. The average LOAS levels range from 15.8 bp for GN30 to 23.2 bp for FH30. As mentioned in Levin [2001], the current-coupon LOAS dynamics is best described mathematically, by a jump-diffusion model with a strong mean reversion. In such a model, jumps simulate rare random crises, whereas the diffusion accounts for day-to-day change in continuous market factors (including the curve shape).
In particular, the quick curve's steepening in the beginning of 2001 may be responsible for the LOAS widening due to sharply injected turnover risk (read above). The new LOAS range of 25 - 35 bps for the conventionals persisted throughout the last few years and even propelled to a higher level of 35 - 40 bps when the accounting scandal broke in mid 2003. This scare, along with refinancing wave, is cooled off and investor confidence is restored, as the LOAS tightening indicates.
Ginnies are usually priced more tightly than Fannies and Freddies, but not always. And, as we mentioned in the December 2003 pipeline, the steep curve is value-detrimental to Ginnies that prepay slower.
As we predicted, Figure 1 exhibits LOAS directionality. LOAS for the conventionals have a negative 66% - 69% correlation with respective mortgage indices (-38% for Ginnies).
Current Coupon MBS Versus Agency Debentures
Taking a direct measurement of OAS off agency debenture curve seems attractive, but requires caution. First, the agency debenture universe is rather sparse and the rates need interpolation to match commonly used maturities. Second, debentures can be issued with drastically differing volumes, raising concerns that points of the curve differ in liquidity.
Figure 2
Figure 2 explores an alternative method: a simple subtraction of a single agency debenture spread (based on the 7-yr maturity) from the FNCL current-coupon LOAS. The resultant spread is much less directional than LOAS as it carries only a -18% correlation with the mortgage rate level. Objective profitability for the government-sponsored enterprises is apparently very low at the end of 2003 and may be compensated only by taking advantage of the ultra-steep curve and shifting the funding risk. Public reports (Bloomberg) indicate that both Fannie and Freddie are shrinking their MBS portfolios.
References
A. Levin, "Mortgage Spread Dynamics", in F.Fabozzi (ed.), Professional Perspectives in Fixed Income Portfolio Management, FJF Associates, 2001, pp. 231 - 239.
R. Perli and B. Sack, "Does Mortgage Hedging Amplify Movements in Long-Term Interest Rates?" The Journal of Fixed Income, December 2003, pp. 7 - 18.
