A focus on voluntary and involuntary prepayment research and application
      

      The MIAC Collateral Behavior Assumption Committee continuously updates the internal voluntary and involuntary vector assumptions to represent market observed performance and anticipated consensus for pricing.  In today’s market, the tightening of credit matrices in new origination programs is the most important attribute effecting loan balance reduction.  The mortgage market became accustomed to very short loan level average lives due to the ease with which borrowers could get credit and quickly take on new or repay old loans.  Prior to 2002, an investment in real estate and taking out a mortgage loan was a well thought out process.  During the five years to follow borrowers would buy homes and incur mortgage debt in very casual manner by historical standards.  The attitude was that refinancing would always be available at a later date to more adequately align funding with homeownership plans.  In August of 2008, where many refinancing options are unavailable, the poor delinquency performance of many loans proves that borrowers we’re not truly committed to honoring their debt for the stated maturity and placed too much dependence on easy credit availability in the future.  As a result, MIAC’s Collateral Behavior Assumption Committee has structured its performance vectors segregated along those lines of new origination availability.

      MIAC has observed client portfolios having voluntary prepayment in the 5% CPR to 8% CPR range for highly levered and or low documentation loans with age of less than 18 months.  The reasoning for the slow speed is simply that these loans have been exposed to the most collateral depreciation coupled with the higher leverage at reduced documentation options that were available in credit matrices at the time of origination.  Although most of these loans were underwritten to Subprime and Alt-A guidelines, many Jumbo program loans have the same problem.  A modeler today cannot just group slower speeds to origination program but must forecast based upon LTV and doc type cut-offs.  Another interesting observation is how documentation type seems to prohibit fast run-off of the older portfolios that still have embedded positive equity on underlying collateral.  Loans with doc types of no income without asset or stated income with stated asset have difficulty refinancing because these programs are mostly extinct in new origination and the job market and economy have not been strong enough to change these borrowers to full documentation credentials.

      Clearly the group with quickest observed voluntary run-off rate in today’s environment are the full doc loans with an adjusted LTV of 80% or less and credit score north of 740.  These are the borrowers that can take advantage of lower rates as a result of the federal rate cuts that have enabled great loan pricing on certain areas of credit matrices offered by the GSE programs and some of the bank portfolio lenders.  CPR for borrowers in the described situation can range from 12% on newer origination paper to 22% on seasoned loans.

      Given the segregation of borrower prepayment ability, MIAC has structured its voluntary prepayment vectors based upon loan to value (LTV) ratio, documentation type and loan size.  In the past, loan structural attributes such as the initial reset on a hybrid ARM would largely drive a modeler’s prediction of run-off, however, in today’s environment vectors must be shaped based on the borrower’s ability to repay the debt instead of the likelihood that they will choose to.

      Involuntary prepayment vectors have gained the most scrutiny over the past year.  The biggest surprise in mortgage credit modeling has been the high rate of default by borrowers with strong credit but low home equity.  The performance has proven that the motive for a lot of borrowers was home price speculation.  MIAC has structured its involuntary vectors to separately address the borrower profile that is simply over leveraged and cant afford the debt as well as for the speculative borrower that has choosen to default over owning the depreciating home.  The MIAC valuation model estimates losses independently, in all major credit sectors: Agency, Alt-A, Subprime, Jumbo, Seconds.  Since current delinquency status of the mortgage pool is the best indicator of the future losses, our forward vectors are created by applying roll rates to each delinquency bucket.  Roll rate is the percentage of the loans in each delinquency bucket that will convert to foreclosure over time.

      The MIAC Collateral Behavior Assumption Committee adjusts involuntary prepayments (CDR) regularly taking into account that overall non-agency collateral performance is worsening with vintage.  For example, 2006 deterioration is more rapid than 2005, and 2007 collateral deteriorates at a seasonably adjusted pace higher than 2006 product (Figure 1).  This tendency can be explained at large by the increased CLTVs of the later vintages. Current MIAC loss projections for 2006-2007 vintage are in following ranges;  2006 vintage Alt-A losses ~7-12%, Subprime ~25-35%;  2007 vintage Alt-A losses ~10-15%, Subprime ~35-45%.  Our loss analysis currently assumes -20-25% home price index decrease as a base scenario. The losses also imply that ~60% current subprime borrowers will be in default with 50%-60% severity.

Figure 1. Subprime Cumulative Loss (by origination vintage)

      Collateral data analysis leads to the conclusion that non-agency hybrid products with shortest resets represent the weakest credit characteristics, with lower FICO and higher than average CLTVs.  These products represent the weakest borrowers, who relied frequently on a positive HPA path and low mortgage payments through interest-only and short reset products. Rate resets in 2008-2009 coupled with the inability to refinance into a new mortgage due to home price depreciation and origination shutdown explains most of the poor Subprime sector performance.

      The deterioration in the 2006-2007 Jumbo sector is on the rise with rest of the Non-Agency universe, and as expected, has accelerated for high CLTV.  We project Jumbo losses falling into 1-4% across the 2006-2007 vintage.  Jumbo losses are a small fraction of Subprime and Alt-A, and can be attributed to much better collateral characteristics and more outlets to refinance than for other Non-Agency sectors.  For instance, for 2007 vintage Jumbo fixed products the FICO = 745, CLTV= 76, while for subprime these characteristics are 625 and 81 respectively.

      We have witnessed a worsening of credit performance for some of the asset classes over the last quarter but some stabilization in the SubPrime sector in terms of total delinquencies.  We believe that the trend will continue into 2009 if housing does not stabilize and liquidity fails to return to the Capital Markets.  Home prices continue to drive credit quality and the uncertainty seen in the home price indices will need to show consistent improvement before an uptrend in collateral performance can be priced into mortgage assets.  Defaults for risky products, such as Payment Option Arms will continue to follow an exponential path given the weakening collateral attributes and risk appetite for this product.