Citigroup Global Markets released a new hybrid ARM model - made available on Citi's Yield Book starting last Monday - utilizing the same modular framework as the firm's fixed-rate models. Analysts said that the existing framework is general and flexible enough so that it can be used to obtain prepayment models for all kinds of mortgages.

However, despite applying the general framework to the new ARM component, analysts considered the differences in prepayment characteristics between hybrids relative to fixed-rate collateral in making the hybrid model.

The first six elements considered have to do with seasoning, baseline speeds and the spike hybrids experience at reset. Analysts noted that hybrids season faster and have higher baseline speeds compared to fixed-rates. They also mentioned that peak hybrid speeds are marginally lower during a refi wave and that even well out-of-the-money ARMs appear to experience refinancing activity. Aside from these, there is a spike in hybrid speeds during the first reset date, and after the initial reset, when the hybrid becomes a one-year ARM, there is less responsiveness to interest rate drops.

"These observations are exemplified by individual cohorts as well as on an aggregate basis," Citi analysts pen. "Furthermore, these observations are consistent with the assumed profile of the agency ARM borrower."

Aside from these observations, Citi also notes that ARM speeds are less sensitive to coupon or incentive. This is because hybrid coupons reset to market rates at the next reset date. Thus, ARM borrowers can only save money from refinancing at this date. By contrast, fixed-rate borrowers have until the loan's maturity to get back their refinancing cost.

Sub-models within framework

The new hybrid prepay model is made up of four sub-models: turnover, refinancing, curtailment and default. The last two elements are not as important to ARM prepayments so analysts looked more closely at the turnover and refinancing aspects instead.

In completing the turnover sub-model, Citi analysts examined the higher relative mobility and seasoning in hybrids. They also considered the weaker lock-in effect of ARMs, due to the faster speeds on discount hybrids.

The next most important element is refinancing, which accounts for many of the differences between the hybrid and fixed-rate models.

In terms of refinancing incentive, Citi explains that, similar to the fixed-rate model, the refinancing incentive for ARMs is calculated by comparing the cost of refinancing with the interest expense savings over an assumed borrower horizon. However, in the hybrid model, analysts considered the fact that the coupon would most likely change at reset. Thus borrower savings are computed based on expected monthly coupons throughout the borrower horizon.

Another difference between the models is the refinancing S-Curves are flatter for ARMs. In terms of the effective mortgage rate used to compute the refi incentive for ARMs, it is based on the weighted average of hybrid and 30-year fixed mortgage rates. Analysts explain that ARM refinancings tend to move into another ARM or another 30-year fixed rate mortgage.

In order to capture the effect of low rates on hybrid refis, analysts also assume the existence of an elbow shift prompted by the media effect. A dip in mortgage rates to multiyear lows will create a media effect in the model. This effect leads to an elbow shift that makes incentive for ARM borrowers stronger.

Hybrids experience a surge in speeds during the first reset. The new model assumes a spike in speeds starting several months prior to the first reset and ending several months after this initial reset. Analysts did point out, however, that the intensity of the spike varies, depending somewhat on burnout

Citigroup also considered the dampened responsiveness following the first reset. Although this occurs in very seasoned and burnt-out fixed-rates, it appears to happen sooner for ARMs. The model captures this through a "media effect burnout" that limits the impact of low mortgage rates on hybrids after the initial reset date.

Caveats to the model

Researchers said that the model accurately captures the overall trend of prepay speeds in different scenarios. But, in general, econometric prepay models are designed with a long-term view, so that average prepay behavior in varied interest rate environments can be tracked, as opposed to accurately monitoring month-to-month speeds for all cohorts.

This is particularly applicable to ARMs, as speeds are typically more erratic than those of major fixed-rate cohorts. This is partly due to the smaller cohort sizes and also partly because of the heterogeneous nature of ARMs. There is also the aspect of ARM borrowers themselves being more heterogeneous compared to fixed-rate borrowers. Citi analysts also mentioned that while the model had been calibrated to generic hybrid data, it could also be adapted to show views on growing niches of the hybrid market such as GNMA, Alt-A and IO hybrids.

Copyright 2004 Thomson Media Inc. All Rights Reserved.

Subscribe Now

Access to a full range of industry content, analysis and expert commentary.

30-Day Free Trial

No credit card required. Access coverage of the securitization marketplace, including breaking news updated throughout the day.