Projecting Housing Turnover:

An Important Issue in our Current Economic and Housing Climate, by Michael Bykhovsky, CEO, Applied Financial Technology

As interest rates rise, prepayment due to housing turnover has become more important. There is an unusually wide discrepancy between projections for at- and -of-money collateral among different models. There are several ways to estimate the speeds for this collateral. One is to examine the historical behavior – how fast were mortgages prepaying when they had a similar interest rate incentive. A problem with this approach is that the recent rate of US existing home sales is almost double of what it was 10 years ago. In theory, we should expect the collateral under otherwise similar interest rate incentives to be prepaying almost twice as fast. This is one example of how modeling techniques that blindly consider history may turn out to be just plain blind.

Another technique assumes that prepayments for this type of collateral, being results of existing home sales, are connected to the index of existing home sales in the US on a one-to-one basis. The most recent existing home sales rate has been about 6.5 - 6.9 million units per year. This is an entirely unprecedented rate in terms of absolute numbers and relative to the existing housing stock. The question is what would be a reasonable assumption about the index going forward. Some modelers make certain assumptions about the future state of the economy or home price appreciation rates. As a consequence these modelers make certain predictions about the expected rate of US home sales. Other institutions go as far as to predict that under unchanged interest rate scenarios, the rate of existing home sales will drop to about 4.8 million units per year. This prediction is actually disparate from what most economists currently feel. Their general estimates place it at about 5.8 million units per year. If one were to make a projection of the rate of existing home sales that is about 20% lower than the one AFT makes, their prepayment speed projections would be about 20% lower. Even though any set of assumptions may turn out be right, we do not feel that as prepayment modelers we should place bets on the shape of the economy that is inconsistent with the consensus of the economists.

Economists make certain assumptions about future movements in interest rates. Based on those movements they make housing sales projections. We correct for the interest rates bias to derive what the consensus would be if the interest rates had stayed unchanged. This is how we get the 5.8 million units of home sales per year figure. This projection is placed into a file with the corresponding base line mortgage rate projection. If projected mortgage rates move above the base line we decrease our existing sales projection, slowly creating a mean from back to normal, and vice versa. As projecting housing turnover is an important issue in our current economic and housing climate, techniques should not blindly consider history and must also recognize the current economic reality.