In the shadow of the largest refi wave of all time, MBS analysts are scrambling to recalibrate their prepayment models, sources say, trying to assess whether existing models accurately capture the enormity of the boom.
At the same time, investors are beginning to question whether prepayment speeds quoted from current models adequately reflect borrowers' behavior, since in reality borrowers might wait out the wave right up until the last minute.
Analysts said that prepayment models are currently being pushed to their limit with virtually the entire mortgage universe above par. However, they noted that it is in this kind of environment where good prepayment models give a lot of added value.
"We have had four major refinancing waves in the last 10 years so we've been able to calibrate our models to capture the nuances between these refinancing waves,' said Dale Westhoff, senior managing director at Bear Stearns. "Having that information has been a real advantage as we get through this one."
One assumption that is critical in forecasting prepayments is determining the actual mortgage rate that is being offered to consumers.
Most of the models use a synthetic par calculation which serves as a proxy or a theoretical mortgage rate for what the actual consumer rate is. To get the synthetic par calculation, the synthetic par coupon is calculated and a spread is added.
"Right now because of capacity constraints on lenders, we have seen a gapping out of the synthetic par mortgage rate," said Westhoff.
Prior to Sept. 11, the spread between the theoretical rate and the consumer rate was very close to zero. During the week of the attacks, spreads gapped out about 10 basis points. More recently, with volumes continuuing to ramp up, the difference between the theoretical mortgage rate and the consumer rate has widened to about 20 to 25 basis points.
As of Thursday afternoon last week, the theoritical mortgage rate in Bear Stearn's system was right around 6.50% while the consumer rate was right around 6.75% to 6.70%.
Bear has been adjusting its assumption on the synthetic par mortgage rate to match the current consumer rate for its model.
Borrower rates vs.
When the Fed announced that it was not continuing the 30-year bond auction, both the 10-year yield and the mortgage current coupon fell substantially. However, analysts said that the mortgage rate that lenders were offering did not come down as fast.
"Lenders have to think if it's a very temporary, one-to-10 day phenomenon that the 10-year fell that much or if it is going be around for awhile," said Eknath Belbase, a senior analyst at Andrew Davidson & Co.
If lenders offer the low mortgage current coupon rate - and the origination proces begins - and if that rate suddenly goes up along with 10-year yields, lenders would lose money. This partly explains the current stickiness' of rates, meaning that the actual rate given to consumers is higher.
The question now becomes how prepayment models are linked into mortgage rates versus overall Treasury yields.
In using the current coupon rate as an input to forecast prepayments, "the way that a good model would take this into account is by using some sort of average rate rather than accounting for the lowest possible rate," said Belbase.
However, not all models use the mortgage current coupon as an input to predict prepayments. Some may use a mortgage origination rate but this would entail checking with other lenders or having one's own lending operation. And it would be harder for money managers to gather this kind of information without calling up some mortgage brokers or looking up some Web sites where mortgage rates are being offered.
Some models take the gross rate while other models use the net rate as an input to forecast prepayments. Andrew Davidson uses the net rate because it is the easiest thing to observe and get historical data on.
Meanwhile, Andrew Davidson is considering not using the ratio between the rate borrowers could get and the rate they actually have as a measure of refinancing incentive.
According to Belbase, the ratio measure starts to change as rates reach their lows. For example, if mortgage rates went down to 4.5% or 5%, the ratio measure would start to say very different things about incentive compared to when rates would stay at 7% or 8%
"For our model we are really looking at whether ratio will continue to be a good measure of refinancing incentive or whether we need to think about other measures as rates continue to fall,' said Belbase.
The end of the wave
While market players have been talking about the extent to which current prepayment models are capturing this refi boom, the flipside to this question is to what extent these models are going to capture the end of the wave. Are they going to still be accurate?
"If you look at the world of refinanceable coupons, it's been shrinking," said Belbase.
The data for September indicates that 9s were almost mostly gone and 8.5s were only in the $10 billion range. The 7% coupon, however, remained at $200 billion.
Belbase said that if the refinancing boom goes on, a lot of the production in the higher coupons will be gone.
He added that 5.5s have been issued in substantial quantities recently. For this coupon to become refinanceable, it might take rates to go down to 4.5%. Belbase said that if rates continue to fall, fewer people might care than if they fell from 8% to 6%.
According to experts, some of the models that predict prepayments on the fast side seem to have some peak levels that may not really reflect the borrowers' mindset and actual behavior.
Though borrowers might know that they are in the money for a refinancing incentive, they would not necessarily rush in to refinance. There actually seems to be a trigger effect on the back-end when rates start rising again. This is when the last surge of prepayments occur because borrowers feel that they might otherwise miss out on an opportunity to refinance. Meanwhile, some of them sit on the sidelines.
"When you see peaks speeds of 30-something CPR on 6.5% loans, it's not clear to me that lower rates bring speeds that high," said Steven Point, fixed income portfolio manager at Glenmede Trust Co. "They may sustain for quite a while in the 20s, and then if rates rise and people become in danger of losing a refinance opportunity, a final rush in might occur. But I would question some of the peak speeds that some of these models would have."