By Michael Thompson, global head of the market, credit and risk strategies group at Standard & Poor's
Our analysis begins with what appears to be a simple question. Say you are thinking of investing in a U.S. residential mortgage-backed securities deal backed by a pool of negatively amortizing, adjustable-rate Alt-A mortgages. Would you rather invest in a security whose underlying collateral had a current aggregate default rate of 17.9%, 26%, 40.4%, or 46.1%?
The obvious answer would be the deal with a default rate of 17.9%, and certainly not the one with the highest default rate. But the truth is that this is a trick question with no clear-cut answer. All of these default rates apply to the same U.S. RMBS structure.
The market, credit & risk strategies (MCRS) group at S&P has been looking into the delinquency and default rates associated with a particularly onerous class of U.S. RMBS--the Alt-A negative amortizing adjustable-rate mortgage-based structure--to investigate the different ways that constant default rates are either disclosed or can be calculated when endeavoring to value associated tranches in today's illiquid and opaque RMBS marketplace. Then, by examining the extremely wide gaps that can exist among reported CDRs, calculated CDRs and reported delinquencies, clues begin to emerge regarding the appropriate assumptions investors might make when undertaking intrinsic valuations of not only Alt-A neg-am-backed structures but the spectrum of RMBS structures, including U.K. and European structured transactions.
Perhaps the most important point to make is that measures of CDR based on disclosed liquidated loans are both backward-looking and misleading for reaching intrinsic valuations of RMBS. Much the preferred basis for calculating current CDRs is using data for the latest disclosed monthly new collateral pool default balance. Furthermore, balanced against current CDR intelligence, total delinquencies as a percentage of the performing collateral pool balance can be monitored as a basis for making future CDR assumptions. For example, current calculated CDRs for U.S. Alt-A neg-am RMBS turned lower in March, possibly suggesting that previously concentrated default risk for the asset class may be dissipating.
The analysis presented in this article is based on the performance characteristics of five U.S. Alt-A neg-am ARM-based structures of vintages originated between 2005 and 2007 (designated SAMIII05AR7, BSMFT06AR2, SAMIII07AR6, GMTA06AR3, and AHMIT071). All structure-specific performance data is sourced directly to, or is calculated from, information taken from either the pool performance tab on the ABSXchange application S&P's structured finance platform or from the monthly servicing statements that are housed under the documentation tab on ABSXchange.
Tables 1 and 2 summarize the issue highlighted in the opening paragraph, namely, how large discrepancies exist between deal-specific disclosed default rates and various forms of calculated proxy CDRs for those same deals. The first of the four categories of defaults included in the tables is the CDR disclosed in the monthly servicing statements for these deals, which is actually the liquidated-loans-based CDR statistic defined as annualized beginning scheduled liquidated loan balance/total beginning collateral balance. The second category of default measurements represents servicing statements' total delinquencies, which are aggregated cumulative pool delinquencies as a percentage of total (performing and nonperforming) collateral balances. In the third category, MCRS calculates delinquencies as a percentage of the performing collateral balance to reveal a realistic proxy of potential future CDR as a portion of the healthy performing collateral balance. And the fourth MCRS-calculated CDR is derived by annualizing new monthly disclosed defaults as a percent of the prior month's performing collateral pool balance.
In addition to the BSMFT 2006 deal cited in the opening paragraph, note how all of the deals listed in Table 1 display large differences in proxy CDRs relative to the CDR disclosed in each deal's March 2009 servicing statement. Table 2 illustrates that the AHMIT 2007 deal in particular has a MCRS-calculated CDR that is 19 times larger than the liquidated loan-based CDR disclosed for this deal in its most recent servicing statement. On average, the MCRS-calculated CDR figure is nearly seven times larger than the liquidated loans-based CDR that is disclosed in the March servicing statement documents for these particular structures.
In order to identify the appropriate CDR for valuing these securities, we will begin eliminating the default rates that we believe are either misleading or inappropriate for reaching an intrinsic value. The first to go is the servicing statement liquidated loans-based CDR because it is calculated from the value of loans that are liquidated and removed from the collateral pool following default. The main objections to this CDR measure are twofold. First, the liquidated loans-based CDR is inferior to the actual level of newly defaulted loans because it is based on defaults that may have occurred as long as one year ago or longer, which has very little to do with the mortgages that are defaulting in the current period. Second, in today's severely depressed real estate transaction market, few defaulted homes are actually being resold following default and repossession, so the liquidated loans-based figure is being artificially depressed until the rate of home sales rebounds from multi-decade lows.
The next category of proxy defaults to be eliminated is the total delinquency percentage balance that is also reported in the monthly servicing statement reports from the deal's trustee. This is simply because it is calculated as a percent of the total collateral balance as opposed to the performing collateral balance. We consider the delinquent mortgages essentially to be in limbo, representing potential future defaults, and therefore prefer to calculate potential future defaults as a percent of the performing collateral pool balance figure.
Having eliminated the two less preferable benchmarks for CDR, we will retain the delinquencies-as-a-percent-of-performing-collateral balance and the actual monthly new default rate, as calculated by MCRS, as the preferable default rates for valuation purposes - because both benchmarks add inherent value, but for slightly different reasons. The MCRS-calculated CDR is a real-time annualized current rate of defaults being realized by the structure, while the delinquencies-as-a-percent-of-performing-collateral balance provides a view for potential future defaults. The outright level of these two default proxies and whether they are converging or diverging can tell volumes about the risk profile of a particular RMBS deal.
Charts 1 and 2 illustrate the previously outlined point about evaluating the risk profile of specific RMBS deals by comparing their current calculated CDR with potential future CDRs as represented by total delinquencies. Chart 1 tracks historical calculated CDR, while Chart 2 tracks total delinquencies. Note how the calculated monthly CDRs for the collateral backing these negative amortizing adjustable rate Alt-A RMBS have on average been consistently higher than the total delinquencies associated with these same structures. This suggests that over the last year, or longer, defaults have been concentrated in recent months as a percent of the delinquent collateral universe, and that many homeowners who are current on their mortgage payments may also be simply walking away from their obligations.
These highly elevated monthly CDRs relative to delinquencies highlights a corrosive characteristic that is unique to a highly leveraged negative amortizing mortgage structure following a period during which adjustable mortgage rates are increasing while home prices are simultaneously falling. The end result is that homeowners are becoming increasingly indebted to a property that is losing its value. This observation helps explain why so many seemingly nondelinquent Alt-A neg-am ARM-financed homeowners appear to be abandoning their homes.
Also note how the elevated calculated CDRs in Chart 1 have recently inflected lower in the direction of total delinquencies, suggesting that the two previously diverging benchmarks may have begun a process of converging. Not only would this be good news for the owners of many of the securities associated with these structures, but it would also increase visibility and therefore confidence in future default assumptions should the gap narrow between current observed and expected future defaults.
These data and analysis evaluate only a single class of U.S. RMBS, one that happens to come very close to scraping the bottom of the toxic asset barrel. The focus on Alt-A neg-am ARMs does, however, magnify the importance of adopting the appropriate CDR that comes closest to true intrinsic valuations. Table 3 reveals the calculated intrinsic values for tranches associated with the AHMIT071 deal analyzed to show how the CDR assumption affects net present value calculations.
The default rates used are listed in Table 1, which are the only variable inputs for the valuation. The constant assumed credit inputs are 5% prepayments, 60% loss severity and a 12-month recovery lag for liquidation of underlying collateral following default. The 60% loss severity assumption may appear too severe at first glance, but keep in mind that the negative amortizing mortgages underlying this structure are typically concentrated in highly stressed regional real estate markets such as California, Florida and Nevada. All modeled cash flows are then discounted at the underlying tranche benchmark and associated spread margin for establishing the price of the floating-rate coupon tranches.
The general observation is how tranche valuations show a tendency to decline as you move from left to right away from the liquidated loans-based CDR that is disclosed in the deal's specific March 2009 servicing statement. Also note how two tranches are only able to maintain a "green" valuation in excess of 85% in the scenario that applies servicing statement liquidated loan-based CDR. These originally designated "green" assets actually slip into the "red," less-than-50%, category, under every other CDR scenario.
The obvious message here is that the CDR disclosed in a U.S. Alt-A neg-am RMBS may not be the one that leads to the closest estimation of a tranche's true intrinsic value. As far as the question of the appropriate CDR is concerned, it is MCRS' opinion that the true intrinsic value on a periodic mark-to-market basis is found somewhere between the figure of total delinquencies as a percent of performing balance and the annualized CDR calculated by MCRS, the basis of which is the current monthly reported defaults as a percent of the prior month-ending performing collateral pool balance.