Poles apart

A long-running dispute between the ‘big three’ credit rating agencies concerning the practice of notching refuses to go away. Saskia Scholtes examines the latest contribution to the debate from economics consultancy Nera

The debate over the practice of ‘notching’, or adjusting, the ratings of underlying assets of collateralized debt obligations continues to show little sign of resolution, despite numerous efforts from the rating agencies, sell-side research analysts and economics consultancy Nera (National Economic Research Associates).

In order to rate a collateralized debt obligation (CDO), a rating agency requires each of the securities in the underlying pool of assets to be rated by that agency. However, in an attempt to maintain diversity it is often the case that CDO managers will include assets in the portfolio that have been rated by rival agencies. The practice of notching has evolved to determine what rating should be awarded to assets only rated by rival agencies.

The latest contribution to what has been a long-running dispute over this issue comes from JPMorgan’s CDO group in London, which has sought to establish whether notching is justified at the lowest level of risk: namely, those CDOs containing triple-A and double-A rated ABS tranches. For these ratings, says JPMorgan, reliable data is in abundance, whereas at the mezzanine level, the nature of the available data has made the task of coming to firm conclusions an uphill struggle.

Fitch’s view is that it would normally give the unrated asset the same rating as Standard & Poor’s and Moody’s – in other words, it believes that ratings from the other two agencies are equivalent to its own. But Moody’s and S&P argue that ratings from other agencies are not equivalent and may notch them downwards because, they maintain, those rival ratings have been assigned in a different way.

Other approaches

Although this practice has evolved within the context of rating CDOs, it is also commonly used when previously unrated securities are included in a mutual fund or other pooled investment vehicle. Other methods such as ‘mapping’ or ‘shadow’ ratings can also be used, but these are less feasible where asset-backed securities are concerned.

A shadow rating is typically calculated using various corporate statistics such as the debt-to-equity ratio and amount of debt outstanding. This methodology is not suitable for rating asset-backed security (ABS) deals simply because such securities do not function in the same way as operating companies and therefore the information typically used for a shadow rating is not available.

Mapping is used by the rating agencies to assess an unrated company for, say, an asset management firm that already has an internal ratings database. The rating agency will establish which credits already have both an agency rating and an internal rating, and determine how the ratings relate to one another. For credits without an agency rating, the agency will then ‘map’ the internal rating to an agency rating within a certain level of comfort.

However, Raviv Shtaingos, a structurer in JPMorgan’s European CDO group, says: “Unlike loans, where mapping from an internal rating system can be acceptable to the rating agencies – asset-backed securities don’t have this alternative available because investors do not maintain proprietary rating systems.”

According to Fitch, S&P and Moody’s will typically discount a Fitch-rated security by between two and six notches, if they choose to accept the rating at all. Fitch argues that the practice is being used by the rival agencies to boost market share in structured finance ratings at Fitch’s expense, by forcing issuers to get Moody’s or S&P ratings on all the underlying assets in a CDO.

However, Moody’s and S&P argue that in many instances, they were not chosen to rate an ABS tranche because, had they been, they would have assigned a lower rating than the agencies that were picked. In one study, for example, Moody’s reviewed non-Moody’s rated commercial and residential mortgage-backed securities deals for which it had been approached for a preliminary rating, but which it did not eventually rate. From its analysis, Moody’s concluded that significant differences – equivalent to several notches – existed between its would-be ratings and those actually given by the other agencies that rated the deals.

Moody’s and S&P also contend that the practice of notching is based upon the fundamental differences between the ratings methodologies used by the various agencies, and how these develop over time. It is claimed that as ratings approaches continue to evolve, methodological differences will become even more pronounced. This, together with the difficulty of monitoring non-rated assets, is cited as justification for notching the ratings of these assets downwards.


In response to criticism from Fitch, Moody’s commissioned Nera to conduct an independent study of the performance of structured finance ratings over time. It was hoped that the results would determine whether ratings from the different agencies can be considered equivalent, or whether they are not equivalent and that, by implication, notching is justified.

The study is the first of its kind to include comprehensive data sets from all three rating agencies, and was eagerly awaited by agencies and structured finance professionals alike. But when it was published in November last year, market participants found that the results were disappointingly inconclusive, falling far short of settling the controversy.

And so the debate over notching continues, particularly since over the past few years, CDOs of ABSs have become increasingly popular among investors. Attracted by a product that provides leveraged exposure to a historically stable asset class, in a structure that typically has a high diversity of collateral in the underlying pool, investors have led the charge in the growth of this market.

Furthermore, Edward Cahill, head of European CDOs at JPMorgan, says that the way in which the market is evolving in itself has important implications for the notching issue. To date the ABS re-securitization market has consisted primarily of CDOs of mezzanine ABSs, and the notching rules have been largely developed with these deals in mind. But according to Cahill: “Over the past couple of years, new products have emerged in the form of synthetic CDOs of triple-A and double-A rated ABSs. For these deals in particular, the practice of notching has a significant impact on the universe of collateral that may be used and the resulting average rating. The practice of notching may limit the amount, diversification and quality of deals that can be done.”

In light of this trend, Cahill and his colleagues at JPMorgan have conducted a further review of the Nera data for triple-A and double-A rated ABSs, and found that at this level, the ratings can be considered to be broadly similar. The Nera study showed that a dual-rated security with a triple-A rating from one agency has around a 90% chance of having a triple-A rating from another agency.

Even when Nera examined the data for all securities – both investment and non-investment grade – on which the rating agencies had assigned divergent ratings, the extent of disagreement was low. On average, for the bonds they shared ratings on, Fitch rated 0.6 notches higher than Moody’s, S&P rated 0.1 notches higher than Fitch, and Moody’s rated 0.1 notches higher than S&P. And since Nera’s report shows that ratings differences were greater for speculative-grade than investment-grade securities, Shtaingos and Cahill note that any differences at the triple-A level are minimal.

The Nera report also examines how the magnitude of ratings differences changed with time elapsed since issue, concluding that the magnitude of discrepancies does in fact tend to increase, as claimed by S&P and Moody’s. At first glance, this suggests that the different agency ratings cannot be considered equivalent, but Cahill notes that the Nera report fails to consider CDOs that contain only ABS tranches rated triple-A.

Nevertheless, JPMorgan argues that some conclusions can still be drawn from the report. Nera examines the dispersion of ratings differences over a ten-year period for all securities rated investment grade, and then examines the same set but excluding ABSs with a triple-A rating from two agencies. For the first category, which includes triple-A securities, the mean difference between ratings over time was less than 0.5 notches in 82% of cases. What’s more, the dispersion of ratings for this category was consistently lower than the dispersion of ratings in the second category, from which triple-A rated securities are excluded.

This finding is especially significant in light of the fact that only 25% of the data for all investment-grade securities represents triple-As. Shtaingos says: “What can be seen is that the magnitude of ratings differences over time for dual triple-A rated securities is minimal. The ratings path of a triple-A is almost identical, although there may be a time-lag in a downgrade” At the same time, Shtaingos adds, the result supports the proposition that differences in ratings monitoring and methodology are less significant with respect to dual triple-A rated securities and thus do not require incorporation through notching.

For JPMorgan, its review of Nera’s study at the triple-A level shows that “the practice of notching should be modified for these types of transactions because in the vast majority of cases, had a major rating agency originally rated a tranche that was rated AAA by either of the other two, then that rating is highly likely to have been triple-A.”

However, despite the compelling nature of these results, structured finance professionals suggest that Nera’s findings and JPMorgan’s conclusions are still not sufficiently robust. For one, the Nera study is predominantly based upon analysis of how a rating from one agency moves in relation to a rating from a rival, rather than any measure of how ratings from S&P, Moody’s and Fitch stand up against their own measures of loss or default.

However, Mark Adelson, structured finance analyst at Nomura, points out that none of the rating agencies defines its ratings primarily with respect to transition rates. “Each rating agency defines its ratings as it sees fit,” he says. “Moody’s chooses to define its ratings primarily by reference to expected loss, while S&P and Fitch define their ratings primarily by reference to likelihood of default.”

So while Nera has chosen to compare the performance of triple-A ratings from the three rating agencies by examining their propensity to be upgraded or downgraded, Moody’s would examine performance by measuring the long-term loss rate of triple-A securities compared with other ratings categories, and S&P and Fitch would consider the long-term default rate compared with other ratings. According to Adelson: “Instead of focusing primarily on transition rates, Nera should have taken either a loss or default approach to assessing the performance of ratings from all agencies. We have to take Nera’s conclusions with an extra grain of salt because of their emphasis on transitions.”

While Nera’s results and JPMorgan’s review have shown that empirically, triple-A rated securities perform in a similar way with regard to transition rates, it still remains to be seen whether the corresponding ratings from the different rating agencies can be treated as equivalent against the measures of default and/or expected loss. However, while such an analysis would indeed be valuable, it would require additional data held by the rating agencies and not made available to the market or to Nera.

So while it may indeed be the case, as JPMorgan argues, that notching is not a justifiable practice for triple-A rated securities, structured finance analysts agree that the issue requires further study. Some analysts suggest that this task should now lie at the door of the regulators in view of the potential for anti-competitive practices, together with the fact that regulatory regimes presume that corresponding ratings from different agencies are equivalent.

As Adelson at Nomura says: “If the presumption that ratings are equivalent cannot be analytically supported, entire regulatory regimes might be based more on wishful thinking than on fact.”

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact info@risk.net or view our subscription options here: http://subscriptions.risk.net/subscribe

You are currently unable to copy this content. Please contact info@risk.net to find out more.

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here