Dominion Bond Rating Service has developed a new model, based on historical data of repeat house sales, that estimates a particular property's propensity to lose or gain value, with implications for credit analysis.
Senior Vice President Susan Kulakowski, who devised the model, explained that, "it gives us a reasonable approach for figuring out how much one could lose, so we can set credit enhancements more accurately, to protect investors."
Kulakowski uses nationwide data supplied by Fiserv CSW that includes census, state, country, metropolitan statistical area (communities with high economic and social integration) and ZIP code information. Where sufficient data is available, in large enough markets, Fiserv CSW assembles it into three indices, broken into the least, middle and most expensive tiers.
"We construct indices based on repeat sales, using two or more sale prices for each property," said Fiserv CSW Chief Economist David Stiff. "We then look at all the pairs of prices that span a particular time period and calculate the average change."
The results are superior to median prices and incorporate less "noise" because they are not subject to the volatility produced when sales rise or fall among different house types.
Kulakowski then applies the index results in a formula that determines the magnitude of credit risk according to the property price tier. She focuses on anything either 85% below, or 175% above, a "typical" value, assuming those properties at either end of the spectrum are most likely to suffer defaults. Why? At the cheaper end, upkeep is more expensive for owners at the margin. At the higher end, homes may be more illiquid.
"Will buyers want to pay the extra for that view, or the wine cellar?" Kulakowski asked.
Since RMBS pools are separated by ZIP code, Kulakowski can match the ZIP codes to any available indices for MSAs or counties and measure the relative expense of the securitized property against its MSA or county. Finally, she adjusts a penalty to reflect additional credit enhancement requirements for those properties at the cheap or expensive extremes. "There is always a data lag, but at least the model tells us how easy it will be to liquidate," Kulakowski said. As prices move, the model adjusts.
"We are forecasting actual declines in the second half of 2006 in all the overvalued markets, including California, Florida, New York, Boston and Washington D.C.," Fiserv's Stiff said.
He added that inventories of unsold homes have already jumped, an event that often precedes price declines by six to nine months and provides evidence that a downturn is starting.
Specific markets may represent even more risk, Stiff warned. In Detroit, home appreciation has only run at 2% over the past six years. In a weak local economy there, with potential layoffs, "we would expect to see more frequent and more severe defaults," Stiff said.
When prices increase slowly, homeowners possess less cushion of net equity to encourage them not to default.
While Stiff expects a price correction in the near future, he does not believe it will be as bad as the previous downturn, which coincided with a recession. The rest of the economy appears to be growing, which puts a floor under housing price declines.
Celia Chen, director of housing economics at Moody's/Economy.com, also predicts a possible slowdown by the end of 2006.
"It isn't showing up yet, but [it] may ... within the next 12 months ...," she said.
Chen forecasted that areas such as Los Angeles could drop 2% in 2007, taking a couple of years to unwind. Florida could fall 2% and Las Vegas about 3% in 2008.
Another important factor to consider is the mortgage industry has flooded the market with affordability products like interest-only loans, which could introduce interest rate and payment uncertainty. Chen does not take that variable into account in her forecasts yet, "since we have so little data."
As long as owners see stability in home prices, they can usually refinance to protect themselves. If however, they could not keep up, an increase in defaults would ensue, compounding a slowdown.
As prices move, the DBRS model will reflect the changes, albeit modestly. The wholesale aggregate data it is working with tends to be smoother than individual points. If values are depreciating at a rapid rate, "I probably wouldn't be able to update the model quickly enough," Kulakowski said.
A $1million property would still have its 40% penalty when it might deserve a 50% penalty.
What she can provide though, is a more granular level of information that helps to refine credit assessments. Without the additional knowledge, enhancement levels would need to be higher, as analysts would tend to overestimate losses, to create a buffer of protection. The ability to dig down to the country and MSA levels sharpens the focus.
(c) 2006 Asset Securitization Report and SourceMedia, Inc. All Rights Reserved.