Reflecting the hybrid ARM product's increasing importance in the mortgage-backed market, Andrew Davidson & Co., Inc. (AD&Co) has recently launched a new hybrid ARM prepayment model.
Prior to the release of this model, the firm combined its existing balloon model prior to reset with its one-year ARM model post-reset to mimic the performance of hybrid ARMs. This was done because the necessary data to model hybrid ARM prepayment experience was unavailable. However, AD&Co was recently able to get data from CPR-CDR, Inc. with six to seven years history on 3/1, 5/1, 7/1 and 10/1 ARMs to update its models.
The total origination used for the models was over $10 billion for 3/1s and 5/1s, a little less than $5 billion for 7/1s and less than $2 billion for 10/1s. Aside from this, there is a considerable amount of post-reset data just for 3/1 ARMs. However, there was no useful data for post-reset on 7/1s and 10/1s.
"Now that we have a true hybrid ARM prepayment model, we are in a better position to accurately forecast the actual prepayment experience of hybrids," said Rob Landauer, director of business development at the firm. "Once you incorporate that model into an OAS-type valuation framework, you get a much better representation of what the future cash flows will be. These should provide a better OAS and effective-duration calculation."
Another factor that prompted AD&Co. to develop a model specifically for hybrids is the fact that these products have started to take over the conventional ARM sector. According to a report released by the firm, hybrids comprised 71% of the approximately $21 billion of conventional ARM securities outstanding in 2000. The percentage went up in 2001, with hybrids comprising 90% of the roughly $25 billion securitized.
The factors in the new hybrid model are aging, interest-rate based refinancing, burn-out and seasonality. These factors are similar to the factors that are being used for AD&Co's existing balloon and fixed-rate MBS models. However, despite having the same factors, the data shows different performance characteristics for the loans in the hybrid pools than it does for balloons and fixed-rate mortgages.
These differences are evident when considering aging and turnover factors as well as the interest-rate effect factor.
According to a report released by the firm, refinancing incentive determines the number of months needed to reach full seasoning on the aging curve in both their hybrid ARM and fixed-rate MBS models. However, unlike in the fixed-rate MBS aging curve, after the peak is reached and as the initial reset date approaches, the aging curve starts to rise and peak around reset.
This behavioral pattern was first seen in the firm's analysis of balloon loans. But balloons behave differently from ARMS in another respect because of two reasons. Prepayment speeds are not usually as high for hybrids compared to balloons unless the hybrid ARM WAC is a significant premium compared to the balloon current coupon. There is also some interest-rate sensitivity seen in the first reset speeds.
These differences in the way balloon loans behave as balloon date approaches and hybrid ARM behavior close to the initial reset is because unlike hybrid borrowers, balloon holders have no choice but to refinance or come up with a balloon payment at maturity. Hybrid ARM holders, on the other hand, have several options depending on the interest rate environment and the shape of the yield curve.
AD&Co also measured the interest rate effect (measured using the ratio between the borrower coupon on the loan and an average of the last three months of balloon current coupon for that maturity) and found that there is a considerable interest rate effect in hybrids but there is much less of this effect relative to turnover than in conventional fixed-rate loans.
A new version
The hybrid model is part of AD&Co's new prepayment model, v4.3.3. In this new model, the firm took into account the significant acceleration in prepayments that began in mid 2000.
"We felt that there were some structural changes driving prepayments," said Landauer. "We really needed to incorporate the 2000 to 2001 prepayment experience into our model to reflect those structural changes."
The company found that prepayment speeds on newly-originated premiums were far faster than the market had ever experienced.
"The aging curve was really non-existent and once the loans or pools became even slight premiums, they experienced a much faster prepayments than had been experienced before, so we recalibrated our model to reflect that," he said.
And even as the MBS market enters the post-refi stage, this phenomenon will continue. The aging function on newly-originated pools and loans has disappeared. These mortgages will continue to prepay extremely fast because of the current ease with which borrowers could refinance.