Demand for floating-rate securities and protection from extension risk are top on the minds of investors who are uncertain about the full extent of future Fed rate hikes and the impact of the 45 basis points rise in the average monthly 30-year fixed mortgage rate since the beginning of May. We strongly advocate GNMA ARMs to such investors in the context of current pricing. In addition to recent price action, relatively low supply, and underlying borrower characteristics that make ARM mortgages less prone to extension risk than FRMs support value in ARMs.
Recent price performance of TBA GNMA ARMs has caused them to underperform Treasuries and swaps. During the course of the last month, while 2-year Treasuries have declined 1:02+ in price, low coupon TBA GNMA ARMs have declined by more than two points. This price decline has led to substantially wider OASs, which are 20-30 basis points wider than levels that prevailed a month ago. Importantly, OASs on GNMA ARMs are positive to the swap curve, signaling an attractive entry point into the sector.
The empirical price performance of GNMA ARMs is statistically well explained by changes in swap rates with much of the price sensitivity concentrated around the 2-year rate. While the historical fit is good, we observe that recent price declines for GNMA ARMs have been greater than expected relative to changes in 2-year swap rates. Based on this analysis, we believe that GNMA ARMs are about 20-24/32 cheap to theoretical fair value.
The effect of this spread widening allows us to create synthetic floaters at positive spreads to LIBOR. For example, investors purchasing GNMA 6.5% ARMs would earn an actual/360 spread of 12 basis points, with monthly resets and no delay. The ability to parlay the basis risk between the 1-year CMT index on GNMA ARMs and earn a spread over LIBOR makes this an attractive alternative to CMOs. Relatively high caps and the flexibility to create synthetic floaters with even higher caps than illustrated here further allow investors to buy securities in accordance with their rate view.
A highlight of our newly released GNMA ARM prepayment model is the sensitivity of the underlying borrowers to any refinancing incentive. Thus, despite the current increase in fixed mortgage rate, projected rates that bring the ARM borrower rate close to or higher than fixed mortgage rates will induce some borrowers to switch from adjustable to fixed rate financing. For example, borrowers backing a GN 7% TBA ARM security bear a current gross mortgage rate of approximately 7.75%. Assuming no change in the index from current levels, on the first reset date in July 2001, these borrowers will see their rates increase by 1% to 8.75% (projected fully indexed rate of 9.15% based on a 6.4% index level). The natural propensity of ARM borrowers to switch for rate incentive and/or certainty reasons will make some borrowers willing to pay the nominal rise in monthly mortgage costs in order to receive the greater certainty of fixed mortgage rates. This phenomenon leads to front-loaded cash flows on ARM securities relative to fixed-rate mortgages. We observe that based on our current speed projections, during the first five years, investors in ARMs receive 68% of total projected cash flows versus 30% of expected cash flows from fixed-rate mortgages. Moreover, under a parallel 100 basis points shift in interest rates, the variability in projected cash flows is far less on GNMA ARMs than on fixed-rate mortgages, i.e. the ARM is less negatively convex.
This lower negative convexity translates into lower base case convexity costs on GNMA ARMs. Moreover, at similar dollar prices to those for 8% fixed-rate passthroughs, we observe that 7% TBA GNMA ARMs offer similar rates of return in the base case as longer duration passthroughs. The more stable return profile on GNMA ARMs can also be seen in the lower variability in prepayment speeds and average lives relative to the fixed-rate security.
Technical factors further support a trade into the GNMA ARM sector. In contrast to conventional wisdom, the backup in mortgage rates has not led to a marked increase in GNMA ARM production. In fact, we observe that despite fixed mortgage rates at levels last seen in 1995/1996, the dollar volume of GNMA ARM originations and applications volume for this product has failed to return to the robust levels seen in the mid-1990s. Thus, current monthly production around the $1.5-$2 billion level compare to a monthly origination volume of $2-3 billion in 1996.
On May 3, the Treasury Department announced the likelihood of eliminating future 1-year T-bill auctions. While we don't expect a final announcement until Aug. 3, the date of the next quarterly refunding announcement, the elimination of the 1-year T-bill will eventually impact the 1-year Constant Maturity Treasury (CMT) index that is used as a reference to calculate rates on most adjustable-rate mortgages. Constant maturity yields calculated by the Treasury are based on a yield curve constructed from yields on the "most actively" traded issues across various maturities. Once this yield curve is constructed, the level of the 1-year CMT index is determined by plotting the yield that corresponds to exactly a "constant maturity" of 1-year. As such, the level of the CMT index does not necessarily correspond to the rate on the corresponding on-the-run issue since these benchmark securities gradually roll down the maturity spectrum, i.e. only on the auction date does a new 1-year T-bill have a maturity of one year.
Elimination of the 1-year T-bill will lead to two potential options for future calculations of the CMT index. We believe either option will raise the level of the index. Firstly, the "constant maturity" point can be interpolated between the 6-month T-bill and 2-year note. We expect this to raise the index by approximately six basis points. Alternatively, off-the-run issues with a maturity around 1-year could be used as "most actively" traded securities to construct the yield curve. This option will also lead to a potential rise in yields around the 1-year part of the constructed curve raising the level of the index.
