The Uses (and Abuses) of Modelling Adjustments
Modelling adjustments in Internal- ratings Based models of Risk Weights
Europe Economics
Europe Economics
The banking industry and regulators agree that a risk-sensitive prudential capital regime has significant benefits. However, regulators are concerned that the current implementation has created strong incentives for banks to intervene strategically in the determination of risk-weights. Such interventions would affect the calculation of prudential capital requirements, ultimately aimed at limiting the size and, hence, the cost of capital charges, at the expense of the stability of the banking sector.
Within this context, this report takes a closer look at the use of modelling adjustments as a potential tool for strategic interventions in internal ratings-based models. In particular, we have gathered data on the modelling adjustments made by banks and conducted an empirical analysis of these against the hypotheses cited by the critics of Internal Ratings-Based (IRB) models.
The analysis presented here offers an intuitive and easily applicable framework enabling the assessment of the evidence for — or against — the existence of tactical or strategic interference by a bank’s management in the calculation of its risk-weighted assets (RWAs).
We have looked carefully for links between those variables identified in the literature as potential reasons for the gaming of IRB models — such as having a high cost of capital and being capital constrained — and the observed variation in RWA modelling changes.
We have found no evidence for such links. This analysis does not disprove the thought that a bank might engage in such activities — but the finding is wholly inconsistent with the hypothesis that this is common practice. Based on what we have observed, RWA modelling adjustments are exogenous to bank-specific performance measures. This means that the argument put forward by critics of the IRB approach is at present unproven. Supervisory attention on significant variations in model inputs may, in time, build such a case. Pillar 3 disclosures will help identify the reasons behind changes in RWAs. Supervisors looking at unexplained variation in models should also examine how the framework’s implementation varies between different jurisdictions. Equally, increased harmonisation across the Single Market would be welcome. But since our work indicates that the system is far from broken, more drastic action around the IRB at this point looks unjustified.
The analysis presented here offers an intuitive and easily applicable framework enabling the assessment of the evidence for — or against — the existence of tactical or strategic interference by a bank’s management in the calculation of its risk-weighted assets (RWAs).
We have looked carefully for links between those variables identified in the literature as potential reasons for the gaming of IRB models — such as having a high cost of capital and being capital constrained — and the observed variation in RWA modelling changes.
We have found no evidence for such links. This analysis does not disprove the thought that a bank might engage in such activities — but the finding is wholly inconsistent with the hypothesis that this is common practice. Based on what we have observed, RWA modelling adjustments are exogenous to bank-specific performance measures. This means that the argument put forward by critics of the IRB approach is at present unproven. Supervisory attention on significant variations in model inputs may, in time, build such a case. Pillar 3 disclosures will help identify the reasons behind changes in RWAs. Supervisors looking at unexplained variation in models should also examine how the framework’s implementation varies between different jurisdictions. Equally, increased harmonisation across the Single Market would be welcome. But since our work indicates that the system is far from broken, more drastic action around the IRB at this point looks unjustified.
The full report is available at http://afme.eu/WorkArea//DownloadAsset.aspx?id=13943