EBA releases its annual assessment of the consistency of internal model outcomes
Posted: 10 January 2019 | Source: EBA
The European Banking Authority (EBA) published today two reports (here and here) on the consistency of risk weighted assets (RWAs) across all EU institutions authorised to use internal approaches for the calculation of capital requirements. The reports cover credit risk for high and low default portfolios (LDPs and HDPs), as well as market risk. The results confirm previous findings, with the majority of risk-weights (RWs) variability explained by fundamentals. These benchmarking exercises, conducted by the EBA on an annual basis are a fundamental supervisory and convergence tool to address unwarranted inconsistencies and restoring trust in internal models.
The credit risk report examines the different drivers leading to the observed dispersion across banks' models. Most of the results are broadly in line with previous exercises, with 50% of the difference in variability explained by the proportion of defaulted exposures in the portfolio and the portfolio mix. The remaining could be attributed to differences in collateralisation and other institution-specific factors, such as risk strategy and management practices, idiosyncratic portfolio features, modelling assumptions, client structure, as well as supervisory practices. This confirms previous findings that RWA variability can be explained, to a large extent, by looking at some measurable features of institutions' exposures.
For LDPs, the risk weight assessments of institutions on a set of common counterparties have been compared. When substituting the risk weight with that of the median institution, the resulting deviations would generally be below 10%. Furthermore, the variability in estimates has been stable in comparison with the 2017 benchmarking exercise.
For HDPs, the estimated values of probabilities of default (PD) and loss given defaults (LGD) have been compared with observed values, i.e. default rates and loss rates. The report presents evidence that the majority of institutions have conservative estimates, in particular when compared with the observed values for the latest year. In comparison with the 2016 exercise, both default and loss rates have decreased more than PD and LGD estimates in recent years, which is likely to reflect a general improvement in economic conditions.
The competent authorities performed an assessment of the internal models, which have been identified as outliers in this benchmarking exercise. In comparison with previous exercises, their monitoring activities are increasingly noticing issues identified by the EBA benchmarking exercise. The same conclusion holds for institutions' internal validations. This is reassuring and indicates that the increased regulatory and supervisory attention paid to internal models is contributing to the consistency of the RWA of internal models.
Market Risk exercise
Compared to the previous exercise, the 2018 analysis shows a reduction in the dispersion in the initial market valuation (IMV) and risk measures. This improvement was expected and is mainly due to the simplification in the market risk benchmark portfolios. Some variability in the results persists, which mainly stems from different interpretations and heterogeneous market practices adopted by the firms. Some of these issues have been addressed, and the quality of the data has improved.
From a risk factor perspective, interest rate portfolios exhibit a lower level of dispersion than the other asset classes, which is most likely due to the use of more consistent practices and assumptions that are more homogeneous across the banks when modelling interest rate risk. This finding confirms the conclusions drawn in last year's analysis.
In line with the previous exercises, a significant dispersion for all the risk measures is observed. More complex measures such as incremental risk charge (IRC) and all price risk (APR) show a higher level of dispersion.
This report has provided input for competent authorities on areas that may require their further investigation, such as IMV variability for some credit spread products. Supervisors should pay attention to the materiality of risk factors not in VaR and, in particular, not encompassed in the IRC models.