The objective of bank capital requirements is to ensure a bank has a high probability of remaining solvent over a specific horizon. The minimum capital requirement in Basel III of 4.5 percent was calibrated so that banks remain viewed as viable and solvent by creditors and counterparties over a one year horizon 99 percent of the time. The capital conservation buffer (CCB) was set to ensure a bank can withstand a severe recession and remain above the regulatory minimums. Lastly, the GSIB surcharge is designed to lower the probability of insolvency of large systemically important banks even more.
The minimum capital requirement, the capital conservation buffer, and the GSIB surcharge calibrations are based on the assumption that all banks are equally risky and the risk does not change over time or in different circumstances. That is, the calibrations are based on “unconditional” estimates of bank risk. By contrast, stress tests are meant to measure the risk of each specific bank when the test is conducted. The stress tests incorporate inherently the assumption that the risk of each bank is different and that the risk changes over time and depends on economic circumstances. The results are “conditional” estimates of bank risk.
The stress capital buffer proposal released by the Federal Reserve on April 10 can be seen as converting the unconditional calibration of the Basel III point-in-time requirements into a conditional calibration. Instead of ensuring that a bank meeting the requirements has no more than the allowable probability of insolvency on average, the new requirement would ensure that each specific bank at the time of its annual stress test has no more than the allowable probability of insolvency. As shown in the figure below, the proposal accomplishes this objective by replacing the CCB with its stress capital buffer (SCB), the decline in the regulatory capital ratio projected for that bank in its annual stress test, subject to a floor equal to the existing 2.5 percent CCB.
This is a clear improvement. When capital requirements are based on average risk, they create an incentive for banks to choose alternatives whose risk is greater than the average. Moreover, when capital requirements vary with economic conditions, they have the potential, at least, to be counter-cyclical. But for the SCB framework to do a good job putting a bank- and time-specific limit on the likelihood that a bank will become insolvent, its components need to be accurately calibrated. In this regard, both the stress loss component of the new framework, and the GSIB-surcharge component, could be improved significantly. In addition, it is completely arbitrary to floor the SCB at 2.5 percent – the unconditional calibration of the Basel III point-in-time requirements – particularly when it is already being combined with a 4.5 percent minimum and a GSIB surcharge, if applicable.
Stress Capital Buffer Shortcomings
As we have argued and demonstrated in multiple research notes, blog posts, and a working paper, (see here, here, here, and here) the current practice of basing stress losses on each bank’s performance under a single scenario is inconsistent with best practices, inaccurate, and prone to unintended consequences.
First, while the Fed sets capital using a single scenario, a bank could fail for reasons that are not captured by that scenario; thus, the stress test failure rates will appear low relative to the severity of the scenario. Consequently, the Fed inevitably makes the scenario implausibly severe in order to make the test feel sufficiently stringent. Of course, that approach sets the SCB excessively high for the banks that are more exposed to the stresses in that specific scenario, but unintentionally understates the SCB for the banks exposed to risks not featured in supervisory stress scenario. Instead, the Fed should be clear about the likelihood it attaches to the scenarios it designs, as well as the allowable probability of insolvency intended for each bank under the SCB.
Second, another important shortcoming of the current SCB proposal is that it makes capital requirements very volatile. There are two sources of volatility: annual changes to the Fed’s stress scenario and annual revisions to the Fed’s own models used to project capital depletion under stress. Using the results of the past five stress testing cycles, the chart below shows the probability that each bank’s SCB volatility is below a given amount reported in the x-axis. As shown by the dashed blue line, approximately 40 percent of banks would have a SCB volatility greater than 1 percentage point under a parametrization similar to the current SCB proposal. This is quite significant and would complicate even further capital allocation at banks and effectively create the need for an additional 100 basis point capital buffer, on top of all the other buffers.
The Fed and large banks already project losses under multiple scenarios using multiple models. The SCB should be derived using the information in those multiple projections. By weighting those multiple projections, the SCB would be more accurate and less volatile. In particular, the SCB should instead be estimated using the results from the Fed’s projections under the supervisory severely adverse scenario and banks’ own projections under BHC-specific stress scenarios. This approach would also mitigate the problem that loss estimates generated using supervisory models are also less accurate because those rely on “one-size-fits-all” supervisory models. As shown by the green solid line in the chart above, calculating the SCB using an average of the banks’ own projections and the supervisory projections under the severely adverse scenario would already significantly reduce the volatility of the SCB. In practice, we are proposing for the SCB to be an average of supervisory models and scenarios and banks’ own models under the BHC own stress scenario, however results under the latter approach are not publicly available.
With these changes, banks that satisfy the SCB-augmented capital ratios will be closer to having equal probabilities of default prior to the addition of the GSIB surcharge. Not only is that a desirable characteristic, it is a precondition for the surcharge’s derivation.
GSIB Surcharge Shortcomings
The GSIB surcharge is intended to reduce the probability of insolvency of failure of a GSIB so that its expected systemic failure costs (probability of failure times systemic failure costs) is equal to that of a not-quite GSIB. The calculation of the GSIB surcharge depends critically on the mapping between a bank’s capital ratio and its probability of default and on the measure used of the systemic cost of failure. However, as described in an op-ed written by our colleague Jeremy Newell as well as in a TCH research note, these two key components are flawed in several significant ways.
First, the Fed’s estimation of the relationship between capital and probability of failure does not take into account the impact of Basel III liquidity requirements or total loss-absorption capacity requirements. As noted by Brooke et al (2015) and Firestone et al (2017), if major liquidity regulations were in effect during the past financial crisis, banks would have had more liquid assets to sell off during stress periods, reducing their fire sale losses. Also, the introduction of long-term debt requirements reduces moral hazard via the elimination of the too-big to fail subsidy.
Second, the Fed, like the Basel Committee, simply asserts that a “systemic score” derived by adding up indicators of a bank’s importance measures the bank’s cost of failure and there is considerable evidence that the score is, in fact, a poor proxy for failure costs (see, for example, Passmore and von Hafften (2016) and Covas (2017)). Moreover, other regulatory changes designed specifically to reduce failure costs, such as living wills, long-term debt requirements, and the elimination of cross-default clauses are not accounted for in the systemic score.
To address these problems, the Fed could estimate the cost of failure using market-based measures of expected cost such as SRISK, as was done by Passmore and von Hafften (2016) and Covas (2017), as well as using other market-based measures such as ΔCoVar by Adrian et al (2016). Calibration to market-based measures of the anticipated costs of failure should allow the Basel Committee and the Fed to develop a superior, empirically grounded proxy for the cost of failure that should account as well for changes that reduce those costs.
In summary, the SCB proposal has the significant strength of making capital requirements more tailored to each specific bank and each specific set of circumstances, but it can be improved. By combining stress capital buffers derived using a multi-scenario, probabilistic approach, with GSIB surcharges that reflect the impact of post-crisis regulations, the proposal can better achieve its objective: namely to ensure that size of banks’ capital buffers are enough for a bank to endure a severe recession without going below regulatory minimums and at the same time satisfy the equal expected systemic impact for GSIBs. The DFAST stress tests disclosures only allow obtain an approximation of the SCB. Importantly, information on risk-weighted assets in the quarter the minimum regulatory capital ratio under stress is reached is generally not available, unless the bank reaches its minimum level at the end of the stress horizon. For this reason, we are unable to accurately estimate the impact of changes in banks’ balances sheets under stress on its SCB. Also, for this reason the estimates presented in the chart do not floor the SCB at 2.5 percent. The estimates of the SCB are close to those reported publically by some banks.
Disclaimer: The views expressed in this post are those of the author(s) and do not necessarily reflect the position of The Clearing House or its membership, and are not intended to be, and should not be construed as, legal advice of any kind.