Single measures are not enough

Ashish Dev considers the contemporary relevance of three Risk articles from 2002-03

Risk magazine published three articles in 2002 and 2003 on the modelling of tail risk in securitisation tranches. They were Credit risk in asset securitisations: analytical model (Risk May 2002, pages S16-S20) and Coarse-grained CDOs by Michael Pykhtin and Ashish Dev (Risk January 2003, pages 113-116), and Random tranches by Michael Gordy and David Jones (Risk March 2003, pages 78-83). The first two formed the basis of the ratings-based approach (RBA) to capital adequacy for securitisations under Basel II. The last one formed the basis of the Supervisory Formula Approach to capital adequacy for securitisations under Basel II.1 The third approach (Internal Assessment Approach), applicable to certain securitisations, essentially utilises the same table as RBA.

The articles had several important themes in common2:

- Analysis of credit risk in tranches of securitisations requires a different approach from credit risk in traditional bonds and loans.

- A single measure ('rating') is not enough to capture the tail risk in securitisation tranches for all different structures and all different underlying portfolios of collateral.

- In terms of tail risk (as measured by economic capital), a mezzanine tranche can be several times as risky as a similarly rated bond or loan.

The last two have suddenly become especially pertinent in the current market turmoil. In residential mortgage-backed securities (RMBSs) with primarily subprime mortgages as underlying collateral, delinquencies and anticipated losses in the pool have risen to levels that can be considered as tail (not catastrophic) events. To everyone's surprise, not only did the spreads on mezzanine tranches of these RMBSs go sky high but also investment-grade bonds (tranches) sustained principal loss. But considering the results of the Risk articles, this is no surprise at all! If such a tranche happened to be thin (say 1-3% of the total RMBS), a slight increase in losses in the underlying pool can wipe out most of the investors' principal.

There is an argument that for the average investor a single measure is much easier to understand than a multi-dimensional one. But securitisations are complex structures, and risk in complex structures is necessarily complex. To box their risk characteristics into a single measure defeats the purpose of telling the investor in the tranche the prospect of loss in all its manifestations, including tail events.

Rating agency professionals are smart enough not to use the same methodology for rating a securitisation tranche as the one used for rating a bond. But in the end they have to express all their analyses in one single measure - a 'rating'. It is not a question of replacing one 'rating' with another by a different methodology. 'Rating' as it stands today is not to be replaced but should be supplemented by a more comprehensive set of measures for all (including tail) risks. However, the process of 'rating' might be so ingrained that everyone expects a single rating for any fixed-income instrument. Worse still is the implication that a given 'rating' implies the same risk of loss to the investor, irrespective of the instrument being rated - in other words, 'a rating is a rating is a rating'.

The Risk articles showed another important implication, which was lost on most investors in the recent liquidity crisis: the tail risk manifestation in a mezzanine tranche cannot be extrapolated to risk in the senior tranche coming further down the road. While a single 'rating' of a mezzanine tranche may severely understate the tail risk, the corresponding single 'rating' of a senior tranche may sometimes overstate the tail risk to the senior tranche. The result depends on many characteristics of the complex structure. Note that most investors in the securitisation market are AAA senior tranche investors. In the recent market turmoil, these investors clearly felt the fear of loss of principal as manifested by the sharp rise in AAA spreads and their unwillingness to roll over maturing commercial paper. The fault can be traced to the root: one single measure or 'rating' irrespective of the complexity of the structure.

It is not difficult to find institutions, with the ability to leverage, that have invested heavily in securitisation tranches with an investment-grade ratings that earn a spread of say, 150 basis points over Libor, in their quest for margin enhancement. One such instance is the case of the German bank IKB, which was recently the recipient of a bail-out organised by the German financial regulator BaFin. These institutions are not the unsophisticated retail fixed-income investor or even the less sophisticated municipalities. But it did not occur to them that the instrument was a securitisation tranche (or perhaps a CDO-squared tranche) and not a bond and that there might be some difference, especially in extreme situations. While lack of understanding of tail risk in their investments is no real excuse for such institutions, the preponderance of one single measure or 'rating' irrespective of the complexity of the structure may have lulled them into complacency.

The fundamental concept that a single measure such as 'rating' is not enough to capture credit risk in complex structures is applicable to CDOs of asset-backed securities or CDO-squared even more than to more traditional securitisations. It is time to rethink and give up the notion 'a rating (say BBB) is a rating is a rating' irrespective of the structure being rated.

This article refers to credit risk in securitisation only. There are other positive and negative characteristics of securitisation as it relates to market risk and liquidity risk. Therefore, there is all the more reason for a multi-dimensional characterisation of risk in securitisations.

- Ashish Dev is managing director at Promontory Financial. Email: [e-mail]

1. See Capital for asset-backed securities by Vladislav Peretyatkin and William Perraudin, in Structured Credit Products, edited by William Perraudin, published by Risk Books, 2004

2. All through the discussion, I assume that an investor holds the securitisation tranche in a larger portfolio with exposure to many different local markets.