An empirical foundation for calibrating the G-SIB surcharge

BIS Working Papers  |  No 935  | 
31 March 2021
PDF full text
 |  38 pages



During the Great Financial Crisis (GFC) of 2007–09, policymakers intervened to prevent the failure of global systemically important banks (G-SIBs) and to alleviate turmoil in the financial system. Following the GFC, the Basel Committee on Banking Supervision (BCBS) introduced measures to reduce the likelihood and severity of a G-SIB failure in the future. Capital requirements corresponding to measures of systemic importance, along with other post-GFC reforms, increased the going-concern loss absorbency of G-SIBs and improved the resilience of the banking sector. The expected impact framework provides a theoretical foundation for these capital requirements based on systemic importance, which are often referred to as G-SIB surcharges. Our alternative implementation of the expected impact framework has the potential to improve the empirical basis of these surcharges and eliminate uneven incentives for G-SIB growth.


We contribute to the banking regulation literature by introducing enhancements to the current implementation of the G-SIB surcharges that could strengthen its empirical foundation and eliminate cliff effects. First, we use Extreme Value Theory to estimate an explicit probability of default function for G-SIBs. Second, we demonstrate the potential of density-based cluster analysis for calculating the reference bank score necessary for the G-SIB surcharge calculation. Third, we demonstrate that alternative approaches to calibrating loss given default (LGD) could allow for equal treatment of all indicators of systemic importance, including the substitutability indicator. Fourth, we introduce two options for a simple and continuous G-SIB surcharge function that combine parameters from the PD function, LGD function, and reference bank score, with the aim of smoothing incentives to grow across G-SIB scores.


We find that these empirically-based alternative implementations of the expected impact framework would result in minor declines in G-SIB surcharges in the aggregate, but would result in the removal of some of the smaller G-SIBs from the list of G-SIBs. Adopting the "supervisory" surcharge function, which is calibrated to maintain the general level of capital surcharges based on the current supervisory consensus, would result in changes of less than 30 bps in individual G-SIB scores, and in moderate changes in G-SIB surcharges. Adopting a surcharge function that uses CoVaR as a measure of LGD would result in both more significant increases in capital and more significant declines in G-SIB scores and surcharges. These findings suggest that these functions could be used to monitor current G-SIB surcharges, particularly by highlighting gains from the cap on the substitutability score and from cliff effects.


As developed by the BCBS, the expected impact framework is the theoretical foundation for calibrating the capital surcharge applied to global systemically important banks (G-SIB surcharge). This paper describes four improvements to the current implementation of the BCBS expected impact framework. We (i) introduce a theoretically sound and an empirically grounded approach to estimating a probability of default (PD) function; (ii) apply density-based cluster analysis to identify the reference bank for each G-SIB indicator; (iii) recalibrate the systemic loss-given-default (LGD) function that determines G-SIB scores, using both the current system based on supervisory judgment and using an alternative system based on CoVaR; and (iv) derive a continuous capital surcharge function to determine G-SIB capital surcharges.

Our approach would strengthen the empirical and theoretical foundation of the G-SIB surcharge framework. Moreover, the continuous surcharge function would reduce banks' incentive to manage their balance sheets to reduce systemic capital surcharges, mitigate cliff effects, allow for the lifting of the cap on the substitutability score and penalise growth in the category for all G-SIBs. In addition, our two capital surcharge functions might be used to monitor G-SIBs' capital adequacy and distortions induced by G-SIB surcharges.