Salomon Smith Barney Inc. last week said that it is making several significant updates to its primary mortgage prepayment model, which, along with Lehman Brothers' model, is one of the two most important and widely used analytical tools employed by fixed-income investors for option-adjusted spread (OAS) research.
In addition to routine data updates, including factoring in a more pronounced drop in home turnover for the year-end, the fine-tuning of the model will include incorporation of recently released Ginnie Mae geographical information on pools, a slowdown in 15-year prepayment speeds, and a mechanism to mimic the "lender solicitation effect," which keeps speeds on high-premiums from declining as quickly as is expected with the current model.
"Our model did a good job last year, but summer speeds in higher coupons didn't come down as fast as our model predicted," said Lakhbir Hayre, the head of the Salomon mortgage-backed securities department. "Based on conversations with people in the industry, it seemed clear that with advances in technology and with information systems, we are more able to target certain groups of borrowers."
Premium Speeds Deceptive
According to a Salomon prepayment analyst, many similar models on Wall Street have been inaccurate for higher premium coupons, mainly because the speeds had been unexpectedly fast this past summer. Because some of the discount turnover speeds have been faster than expected due to the robust economy, some of the new collateral that has been coming out has seasoned very quickly.
The model change makes higher-coupon mortgage issues appear less attractive relative to lower-coupon securities, on an OAS basis. "In the 30-year sectors, the new model...[boosts] the OAS of discounts and analyzing those of premiums," Salomon analysts said in a report. "This phenomenon was particularly evident in Ginnie Maes."
Therefore, the refinancing seasoning ramp (i.e. transient costs) has been lowered in the new model in order to capture the fast speeds and slower-than-expected declines seen on these coupons in 1999.
In addition to the newer low premiums, Salomon analysts have tweaked the model to put in effect a mechanism which mimics the so-called "lender solicitation effect".
Such an effect kicks in at the tail-end of a refi wave, when refinancing volume goes down. At this point, lenders get desperate to sustain volume; otherwise, they begin to lay people off, Hayre said.
This solicitation effect keeps speeds on high-premiums from declining as quickly as is normally expected. The end result: Mortgage bankers begin to go after the people that are harder to get to refinance.
The new model, therefore, now mimics the post refi-wave behavior of brokers and lenders who start soliciting refinancings from higher-coupon borrowers in an attempt to drum up business because of declining refi volume.
"They seemed to make this change to reflect the fact that once you've gone through a refi cycle, mortgage bankers hire a lot of people," said Robert Calhoun, co-director of research at fixed-income specialists Tattersall Advisory Group. "Suddenly rates rise and the refinancing spigot is cut off, so they build up that infrastructure cost. They go for the scraps and crumbs left over - people with higher interest rate loans but lower loan balances. But once all those are gone, they are willing to go after smaller loan balances with premium, more seasoned coupons.
"In other words, they take the easy stuff first, and then go after the harder stuff."
Will It Affect The Buy-Side?
The Salomon prepayment model is considered one of the most significant models by MBS players because it is used in the Yield Book, a system analytical tool made available to most clients.
However, Calhoun says that the changes in the model will probably not have a huge effect on the overall Salomon index that MBS investors use to gauge how fast securities will be paid.
"Although Salomon is speeding up some of the premiums, the only coupons that it will have the greatest effect on are very small cohorts," Calhoun noted. "The old OAS numbers reflected the fact that people were already trading some securities cheaper than what the numbers used in their models had predicted. So basically, the model was not good for some securities. But, of course, some investors might start looking at different things now, since this model will be spitting out different OAS numbers."
However, Calhoun noted that prepayment models are most key for derivative transactions only, because that is when prepayment effects are most magnified. His company, for instance, mainly trades more conservative tranches, so a change in a prepayment model will only have a minimal effect.
"Also, when you have so much of the market below par, prepayment models are less critical," he said. "When you look at discounts, prepayments may be off a couple of CPR [constant prepayment rate]. But it only becomes critical when the market is over par, which is not the case now. So for now, any errors between what the model says and what is really happening is going to be less critical then when the market is at a premium."
Other changes to the model include a slowing down of the 15-year model, especially for conventionals.
"The 15-year model had been too fast," said Salomon's Hayre. "There has been a demographic change in people who take out 15-year loans - they are older than they were five or six years ago. They are less likely to change homes, and typically don't refinance unless there's a huge drop in rates. They are generally affluent, have good credit, and only refinance if money is a problem."
As for a prepayment forecast going forward, Hayre added that he believes speeds will come down, to be followed by a sharp drop at the end of the year, only to be buoyed in the spring because of seasonal pickups.