By Robert Gillman, Morgan Stanley Dean Witter
Earlier this year Morgan Stanley Dean Witter pointed out that THESIS, a securitization of U.K. student loans, appeared to be performing significantly worse than the original rating agency base case assumptions, as MSDW understood them.
In April, Fitch published a surveillance update on the U.K. student loans transactions, affirming the ratings on all classes of THESIS and HONOURS 2. Fitch appears to have based their ratings reaffirmation on three factors:
A recently improving trend of losses based on an expected front-loaded loss pattern, caused by the changes of lifestyle, and fluid personal circumstances students experience following graduation, together with the expectation that borrowers unwilling to pay, rather than unable to pay, will be more apparent in the early years.
A focus by equity participants and the Student Loans Company on improving servicing performance.
The deferment mechanism which, by enabling unemployed or low-paid borrowers to defer repayment of their loans, should help to cushion the portfolio against adverse economic conditions.
While it will take a number of years for a complete picture to emerge, based on long-term demographic and economic trends, the evidence currently available causes us to continue to harbour significant concerns regarding the performance of the student loans portfolios. Accordingly, we believe that Fitch's recent reaffirmation of ratings on both THESIS and HONOURS may prove to be too optimistic, and we continue to urge investors to carefully examine their positions in these bonds.
Opposing Fitch's three
Based on THESIS data for March 2001, we have constructed the following table relating losses, plus provisions for losses, to loans entering repayment in 1998, 1999 and 2000. We have examined all cohorts within the transaction in an effort to test whether the pattern of losses varies materially between older and newer cohorts for loans entering repayment in any year.
There is no evidence at all that cohorts may perform better when loans entering repayment do so some years after the cohort date. At this point, the group of borrowers entering repayment will be less likely to include individuals just graduating, and therefore subject to the changes to their lifestyles which is attributed by Fitch as a significant cause of losses.
In fact, when we consider the columns for loans entering repayment in 1998 and 1999, precisely the opposite conclusion seems to apply, although we acknowledge that the older cohorts are characterized by progressively smaller data samples. Also, because the majority of the data relates to provision levels, rather than realized loss levels 3 , we are in any case somewhat skeptical concerning the true value of these comparisons because of potential deficiencies with the provisioning methodology.
Concerning the provisioning methodology, readers will note that the column for loans entering repayment in 2000 generally indicates materially lower implied loss levels than for loans entering repayment in 1998 and 1999. The reason for this is simple, and is explained by the fact that these numbers only represent provision levels based on a percentage provision allocation to delinquent loans dependent on the time period for which those loans have been delinquent.
Loans that entered repayment in April 2000 cannot have been delinquent for more than 12 months as at March 2001, and in these circumstances the maximum provision rate applied is 50%. Unless it can be clearly demonstrated that 50% of loans that are delinquent at the 12 month stage will be written off, then the use of provisioning data as a forecasting tool for ultimate losses will be unreliable. Therefore, it is misleading to compare these provision levels with provision levels for loans that have been delinquent for in excess of 12 months, which attract a minimum provisioning factor of 80%.
For the reasons set out above, we believe that Fitch's conclusion that performance is improving, based on their analysis of the 1999 cohort only, may be optimistic. While Fitch correctly identify that, compared to the 1998 cohort, the 1999 cohort shows slightly lower indicative losses for loans that went into repayment at both the cohort date and also one year after the cohort date, we note that the two cohorts are to some extent being assessed on different provisioning bases.
Whilst we are aware that the SLC had problems with the introduction of a new computer system in 1998, there is no clear evidence as yet that performance of 1999 and 2000 cohorts has been materially better than earlier cohorts which may have been more directly exposed to the servicing difficulties.
Also, whilst the equity holder will be naturally eager to ensure the highest possible servicing practices are followed, it is not entirely clear what leverage they may have over the servicer providing that the servicer is meeting the standards required in the servicing agreement. For these reasons, it is difficult to attribute a high level of confidence to any expectation that servicing performance will materially improve.
Deferral process and
Since the transactions commenced, economic circumstances have been regarded as favourable, and yet loss histories have still been high. We regard the deferment mechanism only as a helpful source of protection against further deterioration in difficult economic circumstances.
However, the real question that the presence of the deferment mechanism opens up is to ask why does a portfolio, where the obligation to repay is, to a large extent, means tested' and protects borrowers against unemployment, or never achieving close to average national earnings, generate material levels of losses?
We draw the conclusion that this is a unique asset class where losses will frequently arise from non-economic' channels which may ultimately prove hard to control or forecast. Against this background, we note that the original rating agency expected losses in the region of 6.5-7.0% leaves limited room for error against the credit enhancement of only 8.75% below the BBB notes in THESIS.
Actual realised losses
Because of the difficulties in forecasting performance based on the provisioning mechanisms, we believe that the best way to track loss development is to relate realised losses to loans that have moved into repayment and have been in repayment sufficiently long enough to generate realised losses.
We have used the assumption that all of the losses on cohorts from 1991 to 1997 have arisen on loans that moved into repayment in 1998, although there were GBP105m loans that were already in repayment at the closing of the transaction in 1998. We think this assumption is a reasonable proxy based on the belief that once loans start to successfully repay then they will continue to do so. This assertion is also borne out by the fact that provisions on this part of the portfolio appear to have been historically low since inception of the transaction.
THESIS performance monitoring for April 2001 is now available. This includes the first wave of losses from all loans that entered repayment in 1999. Unfortunately, it is not possible to separately attribute losses between loans entering repayment in 1998 and 1999 (and in other years). Losses claimed rose from GBP16.1m in March 2001 to GBP24.8m in April 2001.
However, we have calculated that, since the beginning of the transaction, loans aggregating GBP229.4m have now gone into repayment and have become eligible for losses (i.e., loans that entered repayment in April 1998 or April 1999). At the same time, through April 2001, losses totaling GBP24.8m have been recorded - an implied loss rate of 10.8%. However, we can expect this rate to rise over the next few months as additional losses are recorded, but the eligible loan pool on which losses can be recorded will be unchanged until April 2002.
We understand that the original assumption by the underwriter was that only 10% of the original pool would ultimately run to cancellation. If we deduct this 10%, plus the GBP105m of loans that were already in repayment prior to 1998, and apply the 10.8% implied loss rate to the balance, then we derive ultimate pool losses of GBP89m, which will virtually equate to the level of credit enhancement under the BBB bonds.
Based on our current long-term estimates, our analysis indicates that losses may come very close to the BBB bond level.
Fitch are more sanguine concerning likely future performance. However, we are unable to satisfy ourselves that the agency's reasoning is founded on solid evidence. Based on our analysis, we believe that significant performance improvements will be required to create a more comfortable breathing space between projected loss levels and the credit enhancement available to the BBB notes.
These are, of course, very long-dated assets and, although it may ultimately be the case that it is too soon to tell' as far as long term performance is concerned, we continue to believe that there is sufficient underlying potential volatility in the performance of these assets to require investors to carefully monitor their positions.