Tirupam Goel, Ulf Lewrick and Aakriti Mathur
Reforms following the 2008 financial crisis have led to significant increases in banks’ capital requirements. A large literature since then has focused on understanding how banks respond to these changes. Our new paper shows that pre-reform profitability is a vital, but often overlooked, driver of banks’ responses. Profitability determines the opportunity cost of shrinking assets, and underpins the ability to generate capital. We develop a stylised model which predicts that a more profitable bank would choose to shrink by less (or grow by more) compared to a less profitable bank in response to higher capital requirements. Combining textual analysis of banks’ annual reports with the assessment of a key too big to fail (TBTF) reform, we show that this prediction holds in practice.
The G-SIB framework
The G-SIB framework is a key element of the TBTF reforms. Out of a sample of large internationally active banks, the framework identifies and labels those banks whose failure would be particularly harmful for the global financial system as global systemically important banks (G-SIBs) (we refer to the remaining banks in the sample as ‘non G-SIBs’). For the purposes of identifying these G-SIBs, the framework relies on a measure of systemic importance called the ‘G-SIB score’, which is defined as the weighted average of banks’ market shares in 12 different financial activities. Then, based on 100 basis point intervals of their scores, G-SIBs are classified into five buckets with increasing capital surcharges.
As raising capital is costly, the design of the framework creates incentives for G-SIBs, but not non G-SIBs, to reduce their systemic footprint. Therefore, the framework is well suited to test the differential impact on banks of raising capital requirements, and to investigate the role of profitability.
We use a newly released data set on the annual assessment exercise conducted by the Basel Committee on Banking Supervision. We adjust the official G-SIB score to make it suitable for our analysis. First, we strip out exchange rate effects. Second, we rebase the official score to 2013 values. This is because the official scores are based on a relative comparison of banks within the sample, which implies, for example, that an increase in the average score of all non G-SIBs would mechanically lead to a decline in the average score of G-SIBs. Our rebasement addresses this mechanical effect of a bank’s actions on other banks’ scores.
Identifying the regulatory treatment date
Identifying when banks begin to respond to regulatory reforms is a key challenge for policy evaluation. Typically, major reforms are announced long before they are gradually phased in. This is true for the G-SIB framework as well. The assessment methodology was first published in 2011 alongside an initial list of G-SIBs. The capital requirement schedule was disclosed in 2012, and phased-in as of 2016, initially only applying to banks designated as G-SIBs in 2014.
We propose a two-step approach based on textual analysis to overcome this challenge and identify the relevant regulatory treatment date. First, we count references to the G‑SIB framework in banks’ annual reports to identify when banks incorporated the regulatory requirement into their strategic planning. Our assessment points to a significant increase in the number of framework-related references by G-SIBs in the lead up to the phase-in of the capital surcharges in 2016 (Chart 1 (top)). This contrasts with the continuous decline observed for the non G-SIBs.
Context-based analysis allows us to further narrow down the regulatory treatment date. We classify each sentence related to the G-SIB framework into four categories, distinguishing between general references to the framework from active capital planning discussions. Our interest is in the latter, ie those sentences that reference how the bank is actually responding to higher capital surcharges. Two examples of such sentences are: ‘In the last year, we took some dramatic actions to reduce our G-SIB capital surcharge (…)’ and ‘Additionally, G-SIB buffers will be included in the hurdle rate’. A steep increase in the share of such action-oriented references points to 2015 as the relevant treatment year.
Empirical strategy
A bank’s optimal response to the G-SIB framework reflects a dynamic cost-benefit analysis: the capital relief from a reduction in systemic importance versus the loss in future revenue. Our hypothesis is that the bank’s profitability at the time when the bank incorporates a regulatory change into its business planning plays a central role in how the bank adjusts.
To assess whether differences in profitability matter, we split our sample into four groups of banks based on the median return on assets (ROA) of the sample banks in the years before the regulatory treatment: the more and less profitable G-SIBs, and the more and less profitable non G-SIBs.
Our estimation strategy benchmarks the score adjustments of G-SIBs against those of Non G-SIBs and controls for potential confounding factors, such as differences in banks’ characteristics as well as economic and regulatory differences across jurisdictions.
While we focus on ROA as our baseline measure given its close link to the fundamental profitability of a bank, our results prove robust to using alternate measures, such as return on equity or risk-adjusted returns. Likewise, we confirm that the results are robust to changes in the sample composition, our adjustment of the official G-SIB scores, use of bank-specific treatment dates, or choice of estimation strategy.
Results
We find that while all four groups of banks increased their scores in parallel during the pre-treatment period (2013–14), the average scores of more and less profitable G-SIBs and non G-SIBs diverged after the treatment. The less profitable G-SIBs decreased their scores substantially (Chart 1 (bottom)). The more profitable G-SIBs, by contrast, continued to raise their scores in line with similarly profitable non G-SIBs.
We assess these trends formally by comparing the less profitable G-SIBs with the less profitable non G-SIBs (and similarly for the more profitable G-SIBs and non G-SIBs). In line with the predictions from our model, our results point to a reduction in the scores of less profitable G-SIBs relative to less profitable non G-SIBs, in a range of 16 to 21 basis points. Furthermore, using a triple interaction of G-SIB designation, profitability, and the regulatory treatment suggests that the wedge in scores between more and less profitable G-SIBs widened by about 31 to 34 basis points after treatment. The estimated effects are economically meaningful as they correspond to about a fifth to a third of the buckets that determine banks’ additional capital requirements. Moreover, we find that the contraction is strongest for those G-SIBs that are not only less profitable but are close to the thresholds that determine the level of capital requirement they face.
Chart 1: Framework references in annual reports (top) and evolution of adjusted scores (bottom)
Notes: Top chart: Occurrences of G-SIB related keywords, averaged across framework-mentioning banks, as a percentage share of the total number of words of each bank’s annual report, with 95% bootstrapped confidence intervals. The solid line represents the 31 G-SIBs designated as such before 2015, while the dashed line represents the 12 largest non G-SIBs based on 2013 data. Bottom chart: Change in average adjusted G-SIB scores, between 2015–18 and 2013–14, for more and less profitable G-SIBs and non G-SIBs. More (less) profitable banks are those with 2013–14 average return on assets greater (lesser) than the sample median.
Policy implications and conclusions
The continued increase in some G-SIBs’ scores following the introduction of capital surcharges raises questions about the overall impact of the framework on financial stability. To shed some light on this aspect, we study the evolution of systemic concentration and market-implied default risks for G-SIBs over the sample period. We find that systemic concentration, based both on the market share of the largest G-SIBs and the commonly used Herfindahl-Hirschmann index, has declined after the introduction of the framework. Moreover, market-implied default risks have fallen for the less profitable G-SIBs, resulting in a significant reduction in their systemic risk contribution – in line with regulatory objectives. However, market perceptions of the more profitable G-SIBs’ default risks have not declined in the post-treatment years despite a notable increase in the banks’ capital ratios and greater reliance on more stable sources of funding (Goel et al (2019)).
To summarise, our main finding is that profitability plays a determining role in shaping banks’ responses to higher capital requirements. Taking account of how it drives the varied adjustments across banks helps to evaluate whether regulatory reforms are meeting their objectives. For instance, the diverging response to regulation by more and less profitable G-SIBs uncovered in our analysis argues in favour of continuous monitoring by supervisors. It also calls for further research on how to design and calibrate the capital surcharges to ensure optimal regulatory traction. This would ideally take into account interactions with complementary G-SIB policies, such as enhanced supervision and resolution regimes, which are beyond the scope of the analysis in this paper.
Tirupam Goel and Ulf Lewrick work for the Bank for International Settlements and Aakriti Mathur works in the Bank’s Prudential Framework Division.
If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.
