As we look at the new year, it is quite clear that the "mother of all refinancing waves" is about to start winding down. After the 30-year conventional current coupon yield hit its all-time low of 5.66% on October 31, an almost completely unanticipated bear market in November and December took the current coupon yield up to 6.74% on December 26, its highest level since mid-July, before settling down into a range of 6.45 to 6.65 in early January. This abrupt bear market in bonds caused the MBAA Refinancing Application Index to plunge from its all-time high of 5534.5 for the week ending November 9 to 1564.4 for the week ending December 21, a 71.7% drop in just six weeks! (It has since further fallen to 1284.8 during Christmas week, but that is a holiday aberration). After cheapening in early December on concerns that Japanese withholding taxes would not exempt agency MBS, GNMAs have bounced back to fair value versus convenionals when it became clear that agency MBS would be exempt. In premiums, however, we prefer GNMA IIs to both GNMA Is and conventionals, for both the cheaper prices (especially in 7.5s) and the servicer diversification that characterize GNMA II pools. We believe that the MBS market is priced slightly on the inexpensive side compared to Treasuries and agencies, and suggest a modest overweight at present spread levels. We also believe that swaption implied volatility remains quite high compared to historical averages, and is likely to decline somewhat in the first half of 2002, which would be a significant positive for the MBS market. At the present time, 15-year passthroughs are only slightly cheap to comparable 30-years, and since they do not currently roll as well, we suggest only a market weight for this market sector. If swaption volatility declines significantly from its current elevated level, and the curve flattens a bit, the 30-year sector is likely to outperform. We believe the sector to overweight is 30-year conventional current and cusp coupons (6%-7%), which have slightly higher OASs than premiums and roll better than GNMAs or 15-years. In the CMO market, we like well-structured 5, 7, and 10-year PACs backed by 30-year 6s and 6.5s, at spreads in the high 120s. In mortgage derivatives, the cheapest sector is long WAM IOs backed by 6s and 6.5s, which would benefit from a possible slowdown in the housing market and become less refinanciable if housing prices decline in a slow-growth, low-rate scenario.
Scorching speeds in Dec.,
but the slowdown cometh
The December prepayment reports that came out in early January showed the largest refinancing volume in the 31-year history of the MBS market. This was not surprising, since the end of October and the beginning of November saw the lowest mortgage rates in 30 years. The 30-year FNMA interpolated current coupon yield closed at 5.66% on October 31, 27 basis points below the low of the 1998 cycle (5.93 on 10/5/98). Primary market rates, as measured by the Freddie Mac Weekly Primary Mortgage Market Survey, also reached a new all-time low for the 30-year history of that series at 6.45% on November 8, just four basis points below the low if the 1998 cycle (6.49% on 10/8/98). The wider recent margin between primary and secondary market mortgage rates reflects a less competitive environment among mortgage originators in the late fall of 2001, given that most originators had pipelines at full capacity and did not need to be aggressive on pricing to attract business. Still, some of last month's CPRs were stunningly fast, including 2000 production Gold 7.5s at 75% CPR and Freddie Trust 209 (a benchmark 7.5% Strip with current balance over $550 million) at 82.4% CPR. These are record speeds for a large MBS cohort (there is over $20 billion outstanding of 2000 production Gold 7.5s) and an actively traded benchmark Trust, respectively. The late 2001 refinancing wave has seen homeowners' exercising their refinancing option even more efficiently than in the 1998-99 refinancing wave, at least for less seasoned pools. A key development of the 2001 refinancing waves has been a widening of speed differences between seasoned and unseasoned premiums, with the latter prepaying faster by a much wider margin than in either the 1992-93 or especially the 1998-99 refinancing waves. For example, Gold 7.5s of 2000, 1997, and 1993 have prepaid over the past six months at 58.9, 40.5, and 30.7 CPR, respectively, while in the six month period ending in February 1999 (the peak refinancing period of the 1998-99 cycle), the then unseasoned 1997 production Gold 7.5s prepaid at 42.5 CPR, the same as 1995 production, with 1992 production just a bit slower with a six-month speed of 39.6% CPR.
A key reason for the increase in unseasoned premium speeds has been the growing percentage of refinancings executed through mortgage brokers, who work on commission. In many cases, 2000 production loans have information retained in the originating broker's PC, and with few homeowners whose mortgages are less than two years old looking to take out equity in a refinancing, most such mortgagors qualify for streamlined refinancing. Mortgage brokers have reduced the mortgagor's labor and inconvenience of the streamlined refinancing process to almost nothing. The rise of the independent mortgage broker has made refinancing even more of a lender-driven rather than borrower-driven process than it was in 1998. Each refinancing wave since the first one in 1985-86 has become progressively more lender-driven, leading to increasing efficiency in the exercise of the mortgagor's prepayment option.
Modestly overweight MBS
In the near term, the decision on asset allocation to MBS depends heavily on market direction and implied volatility of derivative markets. If the bond market rallies substantially, there is likely to be a cheaper entry point for an MBS overweight. At current spread and volatility levels, the MBS market looks fairly valued compared to Treasuries, agencies, and swaps. Significant MBS out performance compared to non-callable spread product almost certainly requires some decline in the current elevated levels of swaption implied volatility. Historically, volatility tends to decline as Federal Open Market Committee (FOMC) eases end and the Fed Funds rate becomes less volatile. We therefore suggest that all but very bullish investors modestly overweight residential mortgages at the present time versus Treasuries and agencies. The best candidates to overweight are conventional current and cusp coupons (6s through 7s), which currently have slightly higher OASs than premiums and tend to roll better than GNMAs or 15-years. The 15-year sector is currently fairly valued to very slightly inexpensive versus the 30-year sector, but benefits less from the volatility decline that we expect. Given the prospect of at least a modest curve flattening in 2002, the time is not right to overweight 15-years.
GNMAs underperformed conventionals in early December on the rumor that a proposed Japanese withholding tax would reduce or even eliminate Japanese buying of agency MBS, which would disproportionately impact demand for GNMAs. When the news that the proposed withholding tax would exempt agency MBS was released, lower coupon (6s-7s) GNMA Is bounced back to approximately fair value compared to conventionals. But in 7.5s, we think GNMAs represent value at just 6+ ticks above FNMAs. While premium GNMA prepayment speeds have significantly lagged comparable conventional speeds in the current refinancing wave, the prepayment advantage of premium GNMAs has eroded somewhat for moderately seasoned and seasoned pools, as certain servicers have taken advantage of the opportunity to purchase delinquent loans out of premium GNMA pools at par. With GNMA IIs trading 13+ ticks behind GNMA Is in the 7.5% coupon, we think that GNMA IIs are the better choice for premium GNMA exposure (the difference in payment date is worth only about 3 ticks). Not only are the GNMA IIs less expensive, but the servicer diversification of large multi-issuer GNMA II pools provides protection against unpleasant issuer-specific prepayment surprises.
In the CMO market, we think that well-structured intermediate average life PACs (i.e,. with PSA bands of 100-250 if backed by 30-year 6s and 100-300 if backed by 6.5s) at spreads to Treasuries in the high 120s represent good value versus collateral for non-roll investors. The slightly higher spreads available on PACs with tighter bands do not represent as good a risk-reward trade-off as better-structured bonds. The much more predictable cash flows from these PACs are available currently without an OAS give-up to collateral, although of course the investor does sacrifice the opportunity to take advantage of dollar-roll opportunities.
In derivatives, IOs have generally appreciated to fair value, or in some cases slightly above fair value during the sell-off of the last two months. An exception to this general pattern is the case of IOs backed by long-WAM 6s and 6.5s, whose expected prepayment speeds have slowed dramatically in the recent sell-off. In evaluating the option cost of these securities, it must be taken into account that a rate decline triggered by a weaker than expected economic recovery in 2002 might well lead to lower housing prices, making it more difficult for mortgagors who have not had time to build up equity to refinance their mortgages. Seasoned Trusts, backed by mortgages with lower LTVs, would prepay significantly faster in such an economic scenario.