IRS systems: an industry update

RMA Journal, The, Sept, 2005 by William Chambers

Internal Ratings Systems necessarily vary from bank to bank, which means that compliance with Basel II relies on each institution designing and validating its own system. This article discusses how challenges to this process are being overcome, allowing systems to reach advanced stages of validation, such as benchmarking. The key to success is for a system to have a built-in capacity to evolve.

Unusually for any piece of regulation, the new Basel II capital adequacy framework is intended to provide greater freedom rather than less. By allowing banks to take much fuller responsibility for their risk assessments and regulatory capital requirements, Basel II has been the catalyst for a global focus on the use of Internal Ratings Systems (IRS) to drive lending decisions and portfolio management. Indeed, many banks not aiming for Advanced-Internal Ratings Based (A-IRB) compliance to Basel II--or in the U.S., not aiming for initial compliance at any level--are nonetheless in possession of sophisticated IRS.

Although there is broad consensus over what an IRS can accomplish--aim for economic best practice in banks' ongoing responses to default and non-recovery risk on new and outstanding facilities--the opposite is the case when it comes to how. Indeed, why should the IRS of a German landesbank, a medium-sized Malaysian bank, and an international bank based in the U.S. look anything like one another when these lenders' core markets may differ widely in terms of capital structures, default histories, the use of collateral, and numerous other variables?

The answer is that they shouldn't necessarily look anything like each other. And, as with politics, designing an IRS is the art of the possible; how a system eventually looks will depend not only on the markets it is intended to address but also on which tools are on hand to contribute to it. Systems with a primarily statistical basis require data to be built and tested--the availability of which may be subject to severe constraint. By the same token, a bank that benefits from the presence of highly experienced credit risk personnel would not seek to marginalize this qualitative input by replacing it with purely statistical measures.

The Burden of Proof

But--and it's a big "but"--any bank aiming for compliance to Basel II must be able to prove its IRS in terms of transparency, consistency, full documentation, and the replicability of ratings decisions--even if it addresses different markets and by different means from the next bank's equally "compliant" system. Many of these criteria are intended to allow the greatest possible degree of results feedback--so the accuracy of the IRS becomes self-reinforcing--and to give the bank, its clients, depositors, peers, and national regulators confidence in how facilities are being rated and priced. Yet they do not in themselves address the question at the heart of IRS readiness: how accurately these internal ratings reflect default and non-recovery risk on the portfolio.

Whereas compliance can be tested in terms of the above criteria, testing the accuracy of results is achievable only by a much broader and more subtle set of measures, referred to collectively as validation. And although validation procedures are subject to certain general standards (for example, that the system review function be separated from the designing function), the Basel Committee intentionally stands back from prescribing validation methods. Even when the Committee's validation subgroup made its first public statement on the subject in February 2005, the report stressed that it was not "a comprehensive survey of all available validation methods and processes," nor did it represent "specific Basel Committee guidance to national regulators or financial institutions." Of the six general principles it did outline, one was that "the bank has primary responsibility for validation" and another was that "there is no single validation method"--both reflecting intended diversity between IRSs.

The Three Stages of Validation

Validation involves three general stages:

1. Developmental evidence, which compares the core assumptions of the IRS in terms of wider experience.

2. Benchmarking against how the same or similar credits have been rated externally.

3. Back-testing internal ratings against actual credit outcomes, which is possible only after a certain number of live defaults have taken place. Back-testing is a process that will remain embedded in an IRS for the duration of its use.

Each of the three stages is applied to the examination of the system itself as well as the process used for assigning the ratings. Although these protocols overlap considerably, there is a rough chronology linking them. The current stage can therefore provide some indicator of how long a system has been in development, if not how "ready" it actually is--after all, these successively rigorous procedures are successively more likely to uncover weakness and send the designers back to the drawing board!