A critical starting point for a bank seeking to develop a useful historic database on credit losses is to have an internal credit rating system and allocate a rating to each exposure. Most bank rating systems have approximately 8–12 different rating grades of which 3–4 are devoted to loans now classified in some way as problem or non-performing loans. Such a rating system is necessary to price the credit risk of an exposure. When individual ratings are reviewed on a relatively frequent basis this should provide a basis to determine changes in overall asset quality. During periods when there is a general deterioration in financial conditions this should show itself as a migration of exposures from better quality to lower quality ratings. There are some broad problems with the use of internal ratings systems.
At a minimum the definitions for individual ratings should capture the following: location (city, state, country), industry classification, type of exposure, type and value of collateral, nature of any guarantees and internal credit rating of guarantor. The original definitions are critical because if definitions are subsequently changed they may make it impossible to compare present data with that from the past. The criteria for assignment to a particular band must be clear-cut. The definitions should be largely quantitative and avoid subjective assessments. The risk arising from the latter is that interpretations are likely to change over time leading to the same problems concerning comparability.
While impossible to confirm on the basis of publicly available information, anecdotal accounts suggests that the ratings of new loans granted tend to cluster around a relatively small number of the ratings available. If this were the case it would imply that banks’ differentiation between exposures from a credit risk perspective is relatively crude.
There is a natural tendency to avoid extremes and “clustering” does little to help the objective of differentiating between companies from a credit risk perspective. These classifications can become the source of much argument when they directly affect the level of expected losses, and hence pricing, and unexpected losses and hence, ultimately, allocation of capital.