In 2023, as part of a Capital Requirements Regulation (CRR3) amendment, the probability of default (PD) input floor will rise from three basis points (bps) to five. While this may not seem like a huge increase, it's likely to have a significant impact on risk capital calculations under Basel III.
The general idea of the input modification is to improve PD consistency and to lessen the variability of banks’ risk-weighted-average (RWA) calculations. But the question now is whether the positives of this amendment will actually outweigh the negatives.
Before delving into this issue, it’s important for risk managers to understand fully how we arrived at this stage.
Roughly four years ago, in December 2017, the Basel Committee on Banking Supervision (BCBS) agreed upon a set of reforms to the Basel regulatory accord. These reforms were intended, in part, to address the high level of RWA variability seen in banks’ internal ratings-based (IRB) modeling approaches for credit risk. (They were also supposed to enhance the risk sensitivity of capital requirements in the standardized approach.)
In 2018, this final set of reforms was adopted by the G20 Finance Ministers and Central Bank Governors, and subsequently integrated with the European Commission’s prudential requirements for risk management.
For the EU, the main objective of this Basel 3 finalization is to strengthen banks’ risk-based framework (particularly with respect to the calculation of RWA), without significantly increasing their overall capital requirements. Placing more focus on ESG risks is a secondary goal of the reforms, and other Basel III objectives include the further harmonization of supervisory tools and the reduction of institutions’ administrative costs for public disclosures.
Impact: Pros and Cons
The increase in the PD input floor has serious ramifications.
It is expected that the PD floor amendments will take effect in 2023, while other requirements (such as the output floor) will be gradually phased in over a time span of five years. (The restructured timeline was influenced by the COVID-19 pandemic, and was created, in part, to give banks and supervisors additional time to properly implement the reforms.)
The application of the PD input floor means that exposures currently carrying a PD of 3bps will be increased to 5 bps. Although this increase of 2 bps seems very small, the impact on the banks’ credit risk capital requirements may be substantial, depending on the volume of exposures currently in the 3bps PD bucket. “As could be expected, the increase in PD floors is found to have a sizeable impact on current A-IRB exposures to banks, ” the EBA states in its “Basel III Reforms: Impact Study and Key Recommendations” report.
This begs the question: why did EU regulators (including the BCBS) raise the PD input floor, since it is not an immaterial change?
In its report,“Finalising Basel III,” the Bundesbank provides a strong overview of the pros and the cons of the PD input floor. The Bundesbank argues that the input floors (for both PD and loss-given default) were especially meant to reduce RWA variability – i.e., to deter banks from calculating diverging credit risk RWA for basically the same portfolios.
The smaller a parameter value, regulators contend, “the greater the number of observations needed to validate that parameter value to a statistically significant degree.” Given the scarcity of defaults in some portfolios, they elaborate, it is then difficult to determine such parameters (i.e., PDs) sufficiently. That is essentially their argument for the necessity of a higher input floor.
This line of reasoning, however, is remarkable, since the portfolio that is the most affected by the PD input floor – revolving retail – typically does not have a scarcity of defaults. Moreover, even if a single bank’s portfolio had a low number of defaults, there are “aggregators” that can collect and aggregate PD data. The aggregators could use that information and provide estimates for low-default portfolios, helping to resolve any PD uncertainty.
Furthermore, even if PD estimation remains problematic, it could have been regulated through the existing Margin of Conservatism (MoC) – a sophisticated framework that prescribes add-ons to parameter estimates (e.g., PDs) if there’s uncertainty in terms of risk drivers or default data. The EBA has, in fact, developed a MoC framework in its guidelines for PD and LGD estimation.
Raising the PD input floor, in short, is too simplistic. In practice, under the revised Basel III (with the increased PD input floor), banks will need to combine their lowest PD grades into a single bucket, thereby reducing the granularity at the lower end of the PD master scale. This will be a strange exercise, since PD models are not only monitored in terms of calibration, power and stability but also in terms of homogeneity/heterogeneity.
The implementation of the latest round of Basel III reforms will begin in 2023. Though there has been much ado about Basel III’s output floor changes, the increase in the PD floor, somewhat surprisingly, attracted much less attention.
As we have discussed, the rise in the PD input floor may have a big impact on the required capital calculations for banks – particularly for their revolving retail portfolios. Moreover, this change provides a perverse incentive for risk-averse banks to move their portfolios away from the lowest risk segments.
It's also difficult to understand the PD input floor change from a methodological perspective. Prior to this reform, partly to ensure that there were real differences in predicted and observed default rates, banks were required to implement a PD bucketing system with heterogeneous classes – but, when the PD input floor increase takes effect, they’ll need to combine their lowest PD buckets.
To resolve PD uncertainty, regulators would be wise to look more closely at smarter alternatives (such as using pool data) for low-default portfolios.
Dr. Marco Folpmers (FRM) is a partner for Financial Risk Management at Deloitte the Netherlands. The author wishes to thank Eelco Schnezler, Director at Deloitte the Netherlands, for reviewing an earlier draft of this contribution.