Innovation has been a constant hallmark of the CDO market over the past five years. In the pursuit of new arbitrage opportunities, market participants have sought to apply "CDO technology" to a wider array of assets than the original core of corporate bonds and loans. In 2000, the first CDO exclusively secured by a portfolio of bank trust preferred securities appeared. These securities present some real advantages from a securitization perspective, including high yields and predictable payment patterns, but they also present some significant challenges from a ratings perspective.
Trust preferred basics
Bank trust preferred securities are debt-like instruments that are typically issued by trusts that are owned by bank holding companies. If they meet certain criteria, these securities are treated as additions to regulatory capital, rather than as leverage against the capital base of the bank. In other words the bank can issue what looks like debt, but is treated as equity by its regulators.
In order to qualify for this favorable treatment, they must be 30-year non-amortizing securities. This is substantially longer than most corporate bonds and loans, which are typically no more than 10 years in term. They also contain an option to defer payment of interest for up to five years, without causing default on the underlying asset.
While the issuer is a trust, the underlying bank is the real economic borrower because the bank is receiving the benefit of the additional capital and is the source of the repayment. These securities are attractive to banks because they provide three major benefits. First, they increase the regulatory capital base of the bank without having to issue new stock. Second, the coupon rates on the trust preferred securities might be less than the return on equity new shareholders might demand. Third, the payment on the security is a tax deductible expense, unlike other forms of equity.
Briefly, the major issues with rating a CDO of trust preferred securities that differ from the ratings issues with other, more standard types of assets are:
*The 30-year non-amortizing term of these assets, which is substantially longer than most other CDO assets.
*Many of the issuers for these assets are unrated or have below investment grade ratings.
*The pools lack diversification by industry and sometimes by obligor.
*There are few existing studies of long-term regional bank default rates.
*Projecting long term bank default rates is not an exact science, especially given that the banking industry has undergone substantial changes over the past 20 years and can be expected to go through further changes in the future.
*These assets may defer without defaulting and there is little evidence that shows the rate at which they defer.
Working with data provided by the Federal Deposit Insurance Corp. (FDIC), Fitch calculated a 30-year cumulative gross default rate for banks. The FDIC listing of all bank interventions undertaken since 1934 formed the basis of the Fitch study. The Fitch study included the following features:
*A subset of the universe of banks that mirrors the eligibility criteria of the trust preferred CDOs completed to date.
*A focus on the most recent 30 years of data, which is also the most volatile.
*Counted defaults on an occurrence basis, rather than a weighted average asset basis to emphasize the impact of smaller banks.
*Counted every FDIC intervention as a default, even if no payment default by the bank occurred.
*Counted each bank in a defaulting bank holding company as a separate default to maximize the number of defaults in the study.
Using this approach, the Fitch study concluded a "base case" expected 30-year cumulative gross default rate for banks that meet the eligibility criteria to be 9.83%. Note that in 1999, Salomon Smith Barney conducted a similar study to the Fitch study and concluded similar results. The similar results of independent studies supports the conclusions of both studies.
Trust preferred CDOs have substantially less industry diversification that standard CDOs. The concern with a lack of industry diversification is that if one bank defaults, then numerous more banks may quickly follow suit.
In addition to their work on bank default rates, Salomon Smith Barney performed some research on how regionality of the banks affects their default correlation. Their research divided the United States into five geographic regions and bank default rates by year. They found, not surprisingly, that inter-regional bank default rates are positively correlated. However, they found that these regions are no more correlated to each other than any other five industries in the United States. The implication is that a bank trust preferred CDO that has a national geographic balance is as diversified as any other CDO containing just five industries.
Any CDO that contains just five industries would not be considered substantially diversified, but it is certainly more diversified than a CDO containing just one industry. The lack of industry diversification is one reason why the required stressed default rates required for trust preferred CDOs are high relative to other transactions. More default "cushion" helps offset correlation risk.
Cash flow scenarios
The Fitch cash flow methodology calls for front- and back-end default stresses. The front end default stress for trust preferred CDOs is similar to the standard approach but triples the length of each leg of the default curve in recognition of the significantly longer term of the trust preferred CDO.
The back end default scenario is more different for trust preferred CDOs than the front end. Trust preferred CDOs rely on excess spread and subordination for credit enhancement. If net losses exceed subordination levels, the difference must be made up by excess spread. If the defaults occur late in the life of the transaction there may not be enough periods left to trap excess spread and pay down the senior notes.
In addition, there is the problem of bank crises. Twice in the past seventy years (the Great Depression and the S&L crisis), there have been severe economic dislocations that have resulted in bank failure rates substantially above the norm. Advances in regulatory practices reduce the risk of future banking crises, but their permanent disappearance cannot be assured. To control for this risk, Fitch has designed a cash flow scenario that is intended to simulate the effect of a late term banking crisis. Defaults for the first twenty years are in line with the "base case" rate. However in the last ten years a high-default banking crisis occurs. Because of its timing, there is less excess spread to cushion the defaults.
There is little empirical evidence that addresses recovery rates on trust preferred securities. In light of this, Fitch has decided to take a conservative posture with respect to recoveries, assuming only 10% as an average recovery rate. This is below the level assumed for most other asset types and should be especially conservative because Fitch has defined default as any deferral or intervention by the FDIC. These events do not necessarily imply any loss of principal, let alone 90%.
From the CDO's perspective, prepayments are a mixed blessing for a number of reasons. First, securities that prepay no longer contribute to excess spread. Since excess spread is an important part of the total credit enhancement package, prepayments create a loss of credit enhancement from excess spread. Second, adverse selection may result from prepayments.
A third problem caused by prepayments is potential ratings volatility caused by the increasing concentration of the pool. Most CDO pools start out with a statistically significant number of assets. The large number of assets enables an actuarial approach to the rating of the pool to be used. However, prepayments reduce the number of assets in the pool to the point where the actuarial analysis may no longer be valid and a "weak link" analysis may be appropriate. The weak link analysis focuses on individual obligors and their concentration levels, rather than on characteristics of the pool as a whole.
For more details on Fitch's approach to rating CDOs of trust preferred securities, please visit Fitch's web site at www.Fitchratings.com for the complete text of the article "Bank Trust Preferred Securities Form New Asset Class for CDOs" dated February 2001. In that article, Fitch shows how to leverage its knowledge base in CDOs and its knowledge base in trust preferred securities to assess long term credit risk in these vehicles. The key is to use the CDO technology as a basis, but to be willing to step outside the normal parameters to construct an entirely new rating paradigm that recognizes the unique risks inherent in trust preferred securities.