S&P Viewpoint

Standard & Poor's rates each CDO tranche or liability to the BDR produced by the cash flow model. This BDR, which indicates the cumulative default rate that each tranche can withstand, is then compared to the scenario default rate (SDR) produced by the CDO Evaluator. If the BDR is higher than the SDR, the CDO tranche or liability is deemed able to withstand the level of default stress at the desired rating category.

Much of the information about deal performance/behavior is contained in the BDRs. Standard & Poor's will differentiate its application of BDRs across rating categories by using a percentile approach. As done for other assumptions in the cash flow model, BDRs are separately stressed for a given liability rating, this approach allows Standard & Poor's to implicitly apply confidence levels by rating category, instead of absolute minimums or outliers. Table 1 summarizes the percentiles applicable to each rating category without interpolating at each notch level.

Table 1 BDR Percentiles by Rating
Tranche rating	Percentile
AAA	5th
AA	10th
A	35th
BBB	50th
BB	60th
B	70th

Table 2 shows the sorted break-even calculations.

Table 2 Sorted Break-Evens (%)
Class	A	B	C
Rating	AAA	AA	A
Default pattern
1	50	45	40
2	48	42	35
3	46	39	30
4	44	36	25
5	42	33	20
6	40	30	15
7	38	27	10
8	36	24	5
9	34	21	—
10	32	18	—
11	30	—	—
12	28	—	—
13	26	—	—
14	24	—	—
15	22	—	—
Break-even calculation
	Rating
	AAA	AA	A
Percentile	5th	10th	35th
Break-even (%)	23.40	20.70	17.25

The great majority of CDO liabilities rated in the 'AAA' and 'AA' categories are not structured to accrue unpaid interest, that is, "pay in kind" by the terms of the transaction. On occasion, the BDR is driven by a de minimis interest payment shortfall scenario that the collateral manager can avoid by liquidating collateral or eligible investments. To avoid meaningless BDRs, Standard & Poor's cash flow modeling assumptions will allow for small, self-limiting interest shortfalls for senior tranches, provided, however, that the criteria for PIKable tranches remains unchanged. It is important to note that the interest shortfalls are self-limiting because at some point, the amount of accrued interest forces a decrease in BDRs or an increase in subordination.

The weighted average of the individual asset maturities in the CDO Evaluator should match the weighted average life (WAL) produced by the cash flow modeling for the CDO to achieve the necessary symmetry between the level of defaults a transaction is exposed to and the amount of excess spread it can be expected to receive.

Essentially, the WAL must be an identical input in both the CDO Evaluator and the cash flow models. The assumed WAL of the assets must take into consideration the initial WAL of the pool, the length of the reinvestment period, and the covenanted WAL used in the transaction documents. The collateral amortization schedules in the cash flow modeling should match the transaction documents. Standard & Poor's scrutinizes asset amortization calculations to prevent transactions from deriving risk-free credit from the mismatch in the WAL used in the CDO Evaluator and the WAL captured by the amortization schedule.

Securities with amortization schedules falling both before and after the end of the reinvestment period will be treated as if each amortization payment were a different asset with the same obligor number and rating, and their individual maturity will be computed according to the methodology described above. The maturities so generated, will be used in both the CDO Evaluator and cash flow modeling. (For your convenience, an amortization calculator referencing these assumptions is available Standard & Poor's CDO Interface at www.cdointerface.com.)

Recovery assumptions used in cash flow CDOs and synthetic (loan only) CLOs are modified. Standard & Poor's will stress recoveries based on the rating of the CDO liability. This will bring our methodology for the corporate assets closer to the framework used for recoveries on ABS/RMBS assets. In addition, collateral managers will be provided with the option to incorporate Standard & Poor's CRRs into CDO transactions.

The bond recovery and loan recovery tables (tables 3 and 4) are similar to the recovery methodology previously used, except that they differentiate recoveries based on the ratings of the CDO liability tranche. For example, a senior secured first-lien loan would have a 56% recovery rate when rating the 'AAA' liabilities and a 70% recovery rate when rating the 'BB' liabilities.

In using these tables, one can effectively create a specific portfolio weighted averaged recovery (WAR) for each rated liability class, and use that specific WAR in the cash flow modeling.

Table 3 Corporate Bond Recoveries For U.S.-Domiciled Obligors
CDO liability rating	Senior secured (%)	Unsecured (%)	Subordinated (%)
AAA	48.00	38.00	19.00
AA	49.00	41.00	19.00
A	50.00	42.00	19.00
BBB	51.00	44.00	19.00
BB	52.00	45.00	19.00
B and CCC	52.00	45.00	19.00

Table 4 Corporate Loan Recoveries For U.S.-Domiciled Obligors
CDO liability rating	Senior secured (%)	Unsecured (%)*	Subordinated (%)*
AAA	56.00	40.00	22.00
AA	60.00	42.00	22.00
A	64.00	44.00	22.00
BBB	67.00	46.00	22.00
BB	70.00	48.00	22.00
B and CCC	70.00	48.00	22.00
*In the case of second-lien loans, the first 15% included in the CDO will be treated as senior unsecured loans and the remainder as subordinated loans.

Standard & Poor's will use the assumptions shown in table 5 and 6 for obligors domiciled in Europe. Given the different jurisdictional environment, knowledge, and experience of the local markets, Standard & Poor's will divide the European countries into three groups:

• Group A: U.K., Ireland, and Netherlands.
• Group B: Belgium, Germany, Austria, Spain, Portugal, Luxembourg, Denmark, Sweden, Norway, and Finland.
• Group C: France, Italy, Greece, and Switzerland.

Table 5 Corporate Bond Recoveries For Europe-Domiciled Obligors (%)
Senior secured bonds	Group A	Group B	Group C
AAA	60.00	48.00	43.00
AA	61.00	49.00	44.00
A	62.00	50.00	45.00
BBB	63.00	51.00	46.00
BB	64.00	52.00	47.00
B and CCC	64.00	52.00	47.00
Senior unsecured bonds	Group A	Group B	Group C
AAA	40.00	38.00	32.00
AA	42.00	41.00	35.00
A	44.00	42.00	36.00
BBB	46.00	44.00	38.00
BB	48.00	45.00	39.00
B and CCC	48.00	45.00	40.00
Subordinated bonds	Group A	Group B	Group C
AAA	18.00	18.00	15.00
AA	18.00	18.00	15.00
A	18.00	18.00	15.00
BBB	18.00	18.00	15.00
BB	18.00	18.00	15.00
B and CCC	18.00	18.00	15.00

Table 6 Corporate Loan Recoveries For Europe-Domiciled Obligors (%)
Senior secured loans	Group A	Group B	Group C
AAA	68.00	56.00	48
AA	73.00	60.00	51
A	78.00	64.00	55
BBB	81.00	67.00	57
BB	85.00	70.00	60
B and CCC	85.00	70.00	60
Mezzanine loans/second lien/senior unsecured loans	Group A	Group B	Group C
AAA	45.00	40.00	35.00
AA	47.00	42.00	37.00
A	50.00	44.00	39.00
BBB	52.00	46.00	40.00
BB	54.00	48.00	42.00
B and CCC	54.00	48.00	42.00
Subordinated loans	Group A	Group B	Group C
AAA	20.00	20.00	17.00
AA	20.00	20.00	17.00
A	20.00	20.00	17.00
BBB	20.00	20.00	17.00
BB	20.00	20.00	17.00
B and CCC	20.00	20.00	17.00

Note that for countries not included in the tables above, Standard & Poor's will assign recovery rates on a case-by-case basis. Additionally, recovery rates will be assigned without interpolating at each notch level.

An additional requirement that will be introduced in the tiering mentioned above will be in the form of updating the documentation to reflect the different recovery tests at different rating levels.

Standard & Poor's CRRs estimate the range of principal likely to be returned to lenders in the event of a borrower payment default (see table 7). The CRRs are based on a fundamental analysis of key factors that Standard & Poor's has concluded are likely to drive post-default recovery (see "Recovery Ratings: A New Window On Recovery Risk," published Sept. 8, 2005, on RatingsDirect). CRRs are benchmarked against the Standard & Poor's Risk Solutions LossStats® Database, which provides recovery data on almost 3,200 defaulted and emerged credits.

Table 7 Recovery Ratings Scale
CRR	Analytical description	Indicative recovery expectation
1+	Highest expectation for full recovery of principal	100% of principal
1	High expectation for full recovery of principal	100% of principal
2	Substantial recovery of principal	80%-100% of principal
3	Meaningful recovery of principal	50%-80% of principal
4	Marginal recovery of principal	25%-50% of principal
5	Negligible recovery of principal	0%-25% of principal

Currently, Standard & Poor's assigns CRRs to secured loans in jurisdictions where it is able to make an assessment of the insolvency of the regime governing defaulted debt, including the U.S., Canada, the U.K and Europe, Mexico, and Australia. In the U.S., Standard & Poor's also assigns recovery ratings to new secured bond issues. CRRs are automatically assigned on every senior secured loan or bond for which Standard & Poor's issues a credit rating.

Table 8 shows the recovery rates for an instrument with a CRR.

Table 8 CRR Recovery Rates (%)
	CRR
	1+	1	2	3	4	5
CDO tranche
AAA	100.00	92.00	84.00	60.00	40.00	16.00
AA	100.00	93.00	86.00	63.00	42.00	17.00
A	100.00	94.00	88.00	65.00	44.00	19.00
BBB	100.00	96.00	90.00	69.00	46.00	21.00
BB	100.00	98.00	92.00	72.00	48.00	23.00
B and CCC	100.00	100.00	94.00	74.00	48.00	24.00

Since the CRR analysis is specific for each instrument, there is no need to differentiate between loans and bonds, etc. Each instrument is assigned a CRR on the same scale.

CDO transactions structured using CRRs must use any CRR available for all assets included in the CDO.

Table 9 summarizes the recovery rates to be used for U.S. CMBS. Recovery rates will be assigned without interpolation at each notch level.

Table 9 CMBS Recovery Rates (%)
	CDO liability rating
	AAA	AA	A	BBB	BB	B	CCC
U.S. CMBS rating
AAA	80.00	85.00	90.00	90.00	90.00	90.00	90.00
AA	70.00	75.00	85.00	90.00	90.00	90.00	90.00
A	60.00	65.00	75.00	85.00	90.00	90.00	90.00
BBB	45.00	50.00	55.00	60.00	65.00	70.00	75.00
BB	35.00	40.00	45.00	45.00	50.00	50.00	50.00
B	20.00	25.00	30.00	35.00	35.00	40.00	40.00
CCC	5.00	5.00	5.00	5.00	5.00	5.00	5.00
NR	0.00	0.00	0.00	0.00	0.00	0.00	0.00
NR—Not rated.

Collateral backing RMBS and HEL securities are often subject to a NWC or AFC. NWC/AFC imposes a ceiling on the incremental interest rate increase and absolute interest rate on each loan. When this ceiling is reached, holders of the securities are contractually due the interest income available. The NWC trigger differs for each loan. In an increasing interest rate environment, NWC/AFC may cause the interest paid on the RMBS certificate to be less than that implied by the stated interest rate.

NWC/AFC concerns are also relevant in the case of synthetic exposures designed to mimic the reference obligations. Synthetic NWC/AFC risk is magnified by the proliferation of CDS assets referencing RMBS reference obligations that employ the ISDA PAUG template. PAUG CDS can be written as fixed cap, variable cap, or no cap, thereby, allowing customizable exposure to NWC. Most PAUG CDS referencing RMBS included in CDOs are written with interest shortfall as one of the floating amount events and fixed cap as the interest shortfall cap basis. But, each PAUG CDS carries different protection premiums.

Analysis to capture the effect of NWC/AFC on the ability of the CDO to meet its payment obligations would ideally drill down to the loan level of the portfolio backing the RMBS and HEL securities. Unfortunately, this approach is not viable. This risk, however, cannot be ignored.

Bounded by a broad approach, Standard & Poor's addresses NWC/AFC concerns at the CDO level by applying some additional stress runs in the cash flow modeling that are based on assumptions backed by empirical data evidencing the characteristics of mortgage loans underwritten between 1999 and 2004. The stresses used are based on past empirical data, which can become out-dated as the interest rate and lending environments change and will be updated from time to time.

Standard & Poor's will look for eligibility criteria that set the upper and lower exposure limits to such securities (prime RMBS, subprime RMBS, prime HEL, subprime HEL). In the absence of such limits, Standard & Poor's will run the analysis assuming a mix of 25% prime and 75% subprime RMBS/HEL securities.

As in the case of cash securities, these synthetics will be subject to the same caps at 12% and 14%, respectively, for prime and subprime RMBS/HELs. An illustration of the impact of AFC on the interest received from cash and synthetic RMBS/HEL securities is shown in table 10.

Table 10 Example Of Impact Of NWC On Cash And Synthetic Securities (%)
LIBOR	6	7	8	9	10	11	12	13	14
Min. W/A spread/premium	5	5	5	5	5	5	5	5	5
LIBOR + min. W/A spread/premium	11	12	13	14	15	16	17	18	19
NWC trigger	12	12	12	12	12	12	12	12	12
Cash asset - interest received	11	12	12	12	12	12	12	12	12
Synthetic asset - interest received	11	12	12	12	12	12	12	13	14
W/A—Weighted average. NWC—Net WAC cap.

Table 10 shows how much interest goes to the holder of a mortgage loan in cash from and as a reference obligation under a PAUG CDS as LIBOR increases. In both cases, the NWC trigger is 12% and the spread on the cash loan and premium on the CDS are 5%. The CDS contract has fixed cap selected as the interest shortfall cap basis.

For mixed cash and synthetic RMBS/HEL asset pools, this analysis should be performed assuming the exposure to synthetic securities is at the midpoint of the maximum and minimum limits for such assets.

Due to the recent popularity of CDO of ABS/MBS transactions where both the assets and liabilities take cash and synthetic form, Standard & Poor's will introduce additional modeling requirements to address some of the risks related to these types of transactions.

Hybrid structures take different forms. These modeling assumptions are to be applied to transactions where the assets and liabilities are cash and synthetic as shown in the chart.

Biasing defaults toward the synthetic assets may produce an additional stress to the transaction due to the enhanced likelihood of drawing on the super senior notes, as well as cause additional potential shortfalls in interest as the LIBOR generating collateral (GIC), which is funding the synthetic is drawn upon at a faster pace. The additional stress may occur when the collateral/GIC spread to LIBOR is assumed to be greater than zero or the super senior notes are drawn to cover the credit protection payment and the spread to LIBOR on the drawn portion of the super senior notes is greater than the spread on the unfunded super senior notes.

The biasing formula used in this case is the following:

Default bias = 2x/(1+x)

(Where x is the initial percentage of synthetic assets or cash assets in a pool.)

For example, if the collateral portfolio has a mix of 30% synthetic assets and 70% cash assets, the applicable default bias would be:

Default bias = 2(0.3)/(1+0.3) = 0.46

In this case, the cash flow model would be adjusted to default 46% of the synthetic assets and 54% of the cash assets, instead of the actual 30%/70% split.

Considering that the documentation for most hybrid transactions provides no statement addressing the issue of having only certain types of assets being referenced as synthetics and others as cash, Standard & Poor's needs to make conservative assumptions whereby the issuer "buys" synthetic assets with a longer WAL than the cash assets and vice versa.

Presently, Standard & Poor's simultaneously applies three amortization curves to all assets. However, a CDO's synthetic assets may be referencing ABS/MBS collateral not subject to prepayments (example, CMBS) while the cash assets may be subject to faster prepayment speeds (example, RMBS) and vice versa. To address this, Standard & Poor's will require that amortization speeds be applied to each sector separately. This means that while the cash assets are run at a fast prepayment speed, the synthetics will be run at a slow prepayment speeds and vice versa.

The risk associated with a maturity bar-belled portfolio could be summarized in the following two cases:

• Case 1: Synthetic assets' WAL is longer than the cash assets. In this case, as the credit events "accumulate" at a higher rate due to the biasing default, the super senior swap will be drawn upon at a higher rate for a lengthier period of time. To the extent the collateral/GIC spread to LIBOR is less than zero, the spread lost from cash amortization will be greater than the spread lost from synthetic amortization resulting in an additional stress to the excess spread in the transaction (assuming the collateral spread to LIBOR on the cash assets is the same as the collateral spread on the synthetic assets).
• Case 2: Synthetic assets' WAL is shorter than the cash assets. In this case, as the synthetic assets amortize, the collateral/GIC is released to repay the liabilities, and in most cases invested in eligible investments, which are assumed to earn LIBOR – 100 since the earnings are not guaranteed. This scenario may be stressful when the collateral/GIC spread to LIBOR is assumed to be greater than zero (assuming the collateral spread to LIBOR on the cash assets is the same as the collateral spread on the synthetic assets).