Faced with declining performance on junior subprime liens, Standard & Poor's has updated its surveillance assumptions for residential mortgage-backed securities backed by closed-end second liens after finding purchase money loans, higher combined loan-to-values and lower credit score borrowers highly correlate with poor performance.
The move comes after an adjustment in ratings assumptions in May and the downgrading of 418 classes of closed-end second lien RMBS on July 19, which totaled $3.8 billion or 6.1% of the $62 billion rated by S&P during the period from January 2005 to January 2007.
The current loss curve for new issuance instituted in May is based upon historical performance of closed-ends and anticipates approximately 30 percent more losses within the first 12 months compared to S&P's original loss curve. While losses historically begin to decrease after 12 months, an updated surveillance curve anticipates progressively greater losses in months 18 to 30, given that the delinquency trend has not shown itself to abate, S&P said in a conference call today.
Compared to prior vintages, the ratings agency has been seeing "anomalous behavior" in closed-end second lien deals post 2004, with "2006 vintages turning out to be the worst performing vintage to date and this vintage has been marked by rapid performance deterioration," an S&P director said in the call.
Losses to date significantly exceed the pre-2005 vintages, which have stayed within a lifetime cumulative loss of 3 percent. As of the most recent June 25, 2007, distribution date, last year's vintage with a 12 month weighted-average seasoning is showing approximately 2.81% cumulative losses, while the 2005 vintage over the same distribution date and 22 month weighted-average seasoning is seeing 4.24% cumulative losses to date.
Last year's vintage has seen a rapid buildup of seriously delinquent loans as well as early payment defaults, the director said. Delinquencies of at least 90 days as a percentage of the current full balance in the 2006 vintage are at 4.71% and for the 2005 vintage equal 5.63%. The second half of 2005 started to see rapid increases in early payment defaults.
Additionally, the worst performing 2006 deals are experiencing high losses, and at least two are experiencing double-digit 90+ delinquencies -- the ACE2006-S02 deal with 15.46% and the New Century2006-S1 transaction with 19.60%. The delinquencies have yet to be transferred into the cumulative loss percentage, with cumulative losses expected to be at around 6 percent for the ACE deal. One commonality amongst the worst performing transactions of last year's vintage is that the preponderance of the defaults are coming from Fremont and Long Beach collateral.
"The vintages pre-2005 were performing to expectation by not exceeding a 2 to 3 percent lifetime cumulative loss, but starting in second half of 2005, we saw an anomalous behavior in 2005 and 2006 vintages with a rapid increase in early payment defaults, especially in 2006 vintages," the director said. "We had expected 2006 vintage to perform slightly worse than previous vintages but not as clearly as it has been performing now."
Loan characteristics of the 2006 vintage are somewhat similar to the 2005 bucket -- with about 39 percent of the 2006 vintage holding FICO scores below 660 and 44 percent of 2005 deals falling below 660, according to the call.
But a closer analysis of the 2004, 2005 and 2006 vintages found that three variables are highly correlated to defaults in the past two years' vintages: FICOs less than 660, CLTVs greater than 95 percent and loans used for purchase money, according to the call. Deals last year with FICOs less than 660 are performing two times worse than the same FICO band in the 2004 vintage and the same trend is visible when looking at 60+ delinquencies for CLTV greater than 95 and purchase money loans.
"These variables are exhibiting anomalous behavior and are disproportionately contributing to defaults," S&P said. Thus, "we've updated our new ratings assumptions with increased default assumptions for these specific variables."
It has increased assumptions with regards to probable defaults by 70 percent for loans of CLTVs greater than 95. For FICOs under 660, the default probabilities will increase depending upon their score. For example, a FICO under 575 will have default probabilities increase 4.5 times while only increasing 1.5 times for FICOs between 600 to 660. Purchase money loans had their probabilities increased 1.15 times.
To further refine its closed-end second lien analysis, S&P will also be incorporating first time borrower data and a description of the terms of the first lien loan underlying the second loan.
Among other things, incorporation of these new assumptions decreases the class size of AAA by 25 percent.
In connection with the 418 rating downgrades last week, S&P also revised ratings or surveillance assumptions focus on current losses to date, losses assumed on currently delinquent loans, and future losses for borrowers who are current on payments. S&P said also adjusted its stress test in consideration of the speed at which the transactions' credit profiles can change.
"The new surveillance assumptions along with the new issue stresses are being implemented now because we believe that losses in the U.S. RMBS backed by closed end second lien collateral will significantly exceed historical precedents and our original assumptions," S&P said.