Christian Stracke outlines five key areas where CDO ratings methodology – and structured finance ratings more generally – need to be improved in order to help restore confidence in the structured finance space.

The dramatic rise in US subprime mortgage foreclosure rates may have been the initial driver of the recent turmoil in global credit markets, but as the crisis has intensified, more and more attention has been paid to the market’s poor understanding of risks embedded in the world of structured finance.

Portions of asset-backed securities (ABSs) – most notoriously those ABSs backed by subprime mortgages but also commercial mortgage-backed securities (CMBSs) and other ABS sectors – that were supposed to carry little to no risk are now not only seen to be very risky, but in many cases are essentially worthless. Blame for the debacle can and should be assigned to every guilty party in the process, including underwriters, investors and borrowers, but special focus should be applied to the credit ratings agencies and their arguably flawed method of rating structured finance deals, and especially collateralised debt obligations (CDOs).

The second wave of panic that gripped the markets in late October and early November 2007, as major banks and brokerages were forced to write down CDO exposures, only further underlined investor concerns with CDO ratings methodologies.

Structured products, especially CDOs, are often referred to as highly sophisticated, mathematically complicated entities, but the core idea of structured products, especially CDOs, is actually quite simple. If you take a number of fixed income assets and put them together, they may all individually have relatively high default probabilities, but, you assume, not all of them will default all at the same time. You then slice the pool of assets into different tranches, and each tranche’s credit quality is determined by where it stands in seniority relative to the rest of the tranches in terms of its claim on the cash flows from that asset pool.

A ‘mezzanine’ tranche, usually further subdivided into tranches ranging from a BBB- rating to an A+ rating, has a claim on those cash flows that is subordinate to the claim enjoyed by the ‘senior’ tranche, which itself is further subdivided into various tranches ranging from AA- to AAA. After the first few losses (taken by the unrated ‘equity’ tranche), additional losses are borne by the BBB- tranche, then even further losses hit the BBB, and so on up the tranche structure.

But, by requiring overcollateralisation, or yields generated by the asset pool in excess of that required to meet the scheduled coupon payments due to the mezzanine and senior tranches, the excess spreads generated each year can offset the defaults that are also expected to occur each year, and everyone wins.

The credit rating of the various tranches of a CDO, then, depend on the ratings agency’s view on three main factors: the amount of defaults the pool of assets will suffer over the life of the deal; the recovery rates on defaulted assets; and the degree to which defaults come in clusters, which is another way of saying the degree to which the assets in the pool referenced by the CDO are correlated with each other.

While not taking issue with the core notion of CDO technology, it can be argued that the rating agencies used flawed methodologies in estimating the number of defaults, the losses given those defaults, and the correlation of those defaults. Those flaws in methodology are one of the primary drivers of the recent credit crisis, as the reputation of the major rating agencies was used to convince investors of the validity of what proved to be invalid risk assessments. Going forward, there are five main areas where CDO ratings methodology – and structured finance ratings more generally – need to be improved in order to help restore confidence in the structured finance space.

Smoothed default rates

Forward expectations of default rates in structured finance are based in part on the sector’s historical default rates. The problem, however, is that historical default rates on structured finance are generally substantially higher than on similarly rated corporates and sovereigns. To get around this inconvenient historical truth, the ratings agencies often claim that the historical default rates on structured finance are abnormally high due to a variety of factors. The blow-ups in asset-backed securities composed of manufactured home loans, aeroplane leases and franchise receivables in the 1990s and early this decade have, among other factors, caused default rates on ABSs and on CDOs of ABSs to be well above default rates on similarly rated corporates and sovereigns.

In order to fit structured finance default rates more closely with default rates on traditional rated credits such as corporates and sovereigns, the agencies generally smooth out the humps in structured finance defaults, effectively excluding the clumps of defaults from the default history.

Ratings agencies should not ignore these defaults in their forward-looking structured finance default rate expectations. Unless and until the ratings agencies conduct a wholesale revision of their structured finance ratings methodologies, CDOs containing ABSs and other structured products should be penalised more severely for holding structured products of a given credit rating than corporates of the same rating.

Correlation problems

Because correlation is such a critical component of structured finance credit ratings, the ratings agencies had a strong incentive to use relatively low correlation estimates in order to gain market share in the ratings-sensitive structured finance space.

Moody’s was particularly bold in its method of measuring correlation between rated corporates, and a chief recommendation for changes in CDO ratings methodology is a more realistic assessment of correlation risk in all sectors and asset classes as the current one seems unrealistic.

The problem all ratings agencies, along with all market participants, must contend with is that actual defaults are relatively few and far between, so the correlation of default – the correlation variable that is most directly relevant in rating structured products – can be difficult to test empirically because of lack of data.

Facing this lack of data, one must find proxies for default correlation. In its analysis, Moody’s compared asset correlation (the correlation in default implied by asset prices) to correlation of changes in its own credit ratings. Unsurprisingly, Moody’s found much lower correlation of credit ratings changes than it did asset correlation. Again unsurprisingly, Moody’s chose to use historical credit ratings correlation in generating its forecasts of future correlation between corporate sectors. Going forward, it is recommended that a much closer look be taken at whether the ratings agencies’ correlation forecasts are realistic assessments of correlation risks, or whether they rely on default correlation proxies which conveniently generate relatively low correlation estimates.

Unreal recovery rates

The problem of how to estimate losses in default is yet another serious concern in CDO credit ratings. In analysing the creditworthiness of a given CDO tranche, the ratings agencies must estimate not only the probability of default on the underlying assets that the CDO holds, but also the recovery rates on those defaulted assets. But estimating recovery rates on ABS products that many ABS CDOs hold – tranches of residential mortgage-backed securities (RMBSs), CMBSs, and other ABS deals – is conceptually extremely difficult.

For a tranche of an RMBS deal, for instance, to default, the cumulative losses of the entire RMBS deal must be sufficient to reach the attachment point (the amount of total losses in the mortgage pool at which the given tranche begins to suffer losses) of that tranche. Because the probability of a default on any investment grade RMBS tranche is, in theory, very low, the expected loss on an RMBS tranche in default should also be very low, if only because a low cumulative loss is in theory more probable than a marginally greater cumulative loss.

Full transparency from the ratings agencies in terms of how they estimate recovery rates on defaulted ABS tranches will be critical for market participants to be able to assess the validity of the credit ratings of CDOs that invest in ABS.

Even while the ratings agencies have been scrambling to adjust their credit ratings methodologies on future subprime RMBS, Alt-A RMBS and CMBS deals, they have not generally applied those methodology changes retroactively. If the methodology changes were to be applied retroactively, they would likely result in another wave of ratings downgrades on existing structured products, including CDOs holding those downgraded securities.

There could be a variety of reasons, including reputational risk and potential liability issues, why the ratings agencies appear eager to keep downgrades to a minimum, especially on senior (rated AAA and AA) tranches, which together often make up 85% to 90% of structured deals. The ratings agencies have therefore argued that their ratings actions on exiting structured products are ‘data dependent’, meaning that they will only downgrade if and when the data on collateral performance confirm market suspicions about the probability of default of rated structured products. This data dependence is, in our view, completely inconsistent with the changes in ratings methodology.

If the ratings agencies now see that their original estimates of foreclosure rates and cumulative losses in the subprime mortgages underlying RMBS were wrong, and are adjusting their estimates with more conservative assumptions for new subprime deals going forward, why are those revised assumptions not being applied to already existing subprime deals? Going forward, market participants are expected to demand that ratings agencies be more consistent in their methodology revisions, applying them both to potential new deals and to already existing deals.

Data access

One of the major problems that has been exposed in the recent credit crisis has been the information asymmetry between what information the ratings agencies can access and what everyone else can access. If, for instance, a bank analyst is trying to gauge the potential for future CDO write-downs at a given bank, knowledge of the structure of the CDOs that that bank has underwritten (and of which the bank has likely retained a portion) is critical.

But the relevant documents describing the structure of those CDOs, their target assets and other issues are generally non-public documents to which the ratings agencies have privileged access because they issue the credit ratings on those CDOs. Put another way, it is very difficult for outside observers to monitor the validity of CDO credit ratings because information about those CDOs is not publicly available.

While many of these problems may seem intractable, the good news is that the investor community now at least understands the overly optimistic nature of CDO ratings methodologies, and over time will put pressure on the ratings agencies to adjust their methodologies to reflect reality. While this pressure on the ratings agencies should eventually set the stage for a revival in CDO and other structured product issuance, getting from here to there will still be painful.

More realistic default, correlation and recovery rate assumptions may raise the cost of capital for borrowers who ultimately, if indirectly, use the structured finance market for funds. Higher costs of capital could prolong the slowdown in structured product issuance, which in turn would prolong the slump in ratings revenues for the ratings agencies.

However, a thorough revision of CDO and other structured product ratings methodologies, together with retroactive ratings changes and a move toward much more transparency in structured product ratings, is critical in order to bring an end to the current crisis in structured finance.

Christian Stracke is a senior strategist with the independent research firm CreditSights.

Moody’s replies

When Moody’s updated its assumptions regarding corporate default and asset correlations for CDOs back in November 2004, it recognised that there are a number of ways to model these correlations that have support from various academics and market practitioners.

After examining the pros and cons of these various methods for deriving correlations, Moody’s chose an approach based on historical rating co-movements using a directional ratings transition matrix (DRTM), which incorporates stresses for regional, vintage and key agent relationships.

Moody’s chose this approach because rating co-movements provide a richer data set than actual defaults, and because ratings are not subject to the same sort of market forces such as liquidity premiums and equity market volatility that might induce spurious correlation in market measures of default probability.

This approach also reflects the fact that Moody’s modelling of CDOs draws default probabilities from underlying asset ratings, which in turn incorporate Moody’s assumptions about the likelihood of default.

These pair-wise asset correlations are not used as a proxy for default correlation, but are a primary input into a Gaussian Copula-based model which also takes each asset’s default probability into account to derive default correlations.

A more detailed discussion of the alternative approaches and the rationale behind Moody’s preferred approach can be found in two papers: “Moody’s Revisits its Assumptions Regarding Corporate Default (and Asset) Correlations for CDOs” (November 2004) and “Moody’s Revisits its Assumptions Regarding Structured Finance Default (and Asset) Correlations for CDOs” (June 2005). As discussed in those papers, Moody’s based its correlation assumptions on rating co-movements because this approach was supported by a reasonably large data set, and because it ensured that Moody’s ratings of derivative instruments such as CDOs was fundamentally consistent with the ratings assigned to the underlying assets.

PLEASE ENTER YOUR DETAILS TO WATCH THIS VIDEO

All fields are mandatory

The Banker is a service from the Financial Times. The Financial Times Ltd takes your privacy seriously.

Choose how you want us to contact you.

Invites and Offers from The Banker

Receive exclusive personalised event invitations, carefully curated offers and promotions from The Banker



For more information about how we use your data, please refer to our privacy and cookie policies.

Terms and conditions

Join our community

The Banker on Twitter