Effect of the GSIB surcharge on the systemic risk posed by the activities of GSIBs

Finance and Economics Discussion Series Federal Reserve Board, Washington, D.C. ISSN 1936-2854 (Print) ISSN 2767-3898 (Online) Effect of the GSIB surcharge on the systemic risk posed by the activities of GSIBs Marco Migueis, Sydney Peirce 2025-029 Please cite this paper as: Migueis, Marco, and Sydney Peirce (2025). “Effect of the GSIB surcharge on the systemic risk posed by the activities of GSIBs,” Finance and Economics Discussion Series 2025-029. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2025.029. NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

Effect of the GSIB surcharge on the systemic risk posed by the activities of GSIBs1 Marco Migueis Federal Reserve Board Sydney Peirce Federal Reserve Board March 12, 2025 Abstract This study assesses whether the introduction of the GSIB surcharge requirement resulted in GSIBs reducing the systemic risk posed by their activities. We find limited evidence of GSIBs managing their activities to avoid increases in their surcharges. For a sample of international banks, proximity to surcharge thresholds is associated to a decrease in the growth of intra-financial system liabilities, underwriting activities, and holdings of trading and available-for-sale securities. In the case of US GSIBs and the method 2 GSIB surcharge, we find some association between proximity to surcharge thresholds and a decrease in the growth of trading and available-for-sale securities and short-term wholesale funding. Keywords: bank capital requirements, banking regulation, GSIB surcharge, systemic risk JEL Classification Codes: G01, G18, G21 1 – Introduction In the evolving landscape of global finance, the resilience of financial institutions is of paramount concern. This is particularly true for global systemically important banks (GSIBs), as their size, interconnectedness, and complexity imply that their failure can have significant repercussions for other economic agents. Recognizing the unique risks posed by GSIBs, the Financial Stability Board and the Basel Committee on Banking Supervision have adopted the “GSIB surcharge” – a capital buffer standard that aims to reduce the negative externalities these firms may pose by increasing their required capital.2 The GSIB surcharge framework measures the systemic risk posed by large banks through a “GSIB score,” which accounts for several quantitative measures of their activity, including 1 The views expressed in this manuscript are ours and do not represent official positions of the Federal Reserve Board or the Federal Reserve System. We thank Ben Ranish and the participants in the Policy Research and Analytics seminar of the Federal Reserve Board for helpful suggestions. 2 Capital buffer requirements aim to increase the resilience of banks by ensuring that they can withstand a degree of losses and remain above with their minimum capital requirements.

indicators of size, interconnectedness, and complexity. Taking the measures used to calculate the GSIB score as appropriate measures of the systemic risk posed by banks, this study assesses whether the introduction of the surcharge requirement resulted in GSIBs reducing the systemic risk posed by their activities. Specifically, we test whether the rate at which a GSIB grows their systemic risk indicators diminishes when their GSIB score is close to the thresholds that result in a surcharge increase. Our analysis provides limited evidence of GSIBs managing their activities to avoid increases in their surcharges. For a sample of international banks, we find some association between proximity to surcharge thresholds and a decrease in the growth of intra-financial system liabilities, underwriting activities, and holdings of trading and available-for-sale (AFS) securities. In the case of US GSIBs and the method 2 GSIB surcharge, we find some association between proximity to surcharge thresholds and a decrease in the growth of trading and AFS securities and short-term wholesale funding. Overall, our results should be interpreted cautiously, given that most of the effects are not robust across specifications and have weak statistical significance. Our study contributes to the literature on the effects of the GSIB surcharge and, particularly, the literature on the impact of the GSIB surcharge on systemic risk and financial stability. Papers in this literature include Goel et at. (2019), who find that GSIBs probability of distress decreased after the introduction of the GSIB surcharge and argue that their systemic importance likely declined; Behn and Schramm (2021), who find that GSIBs lowered their risk taking after being designated as GSIBs; Violon et al. (2020), who find that GSIBs reduce the expansion of their balance sheet after receiving the GSIB designation; Ho et al. 2022, who assess the impact of GSIB designation on the complexity of GSIBs; Behn et al. (2022), who find that GSIBs reduce a range of activities at the end of the year to reduce their GSIB surcharges; Garcia et al. (2023), who also find that GSIBs reduce their balance sheets at the end of a year, particularly their derivatives book and their intra-financial system exposure, to reduce their GSIB surcharge; Bery et al. (2024), who also find that US GSIBs lower their surcharges by decreasing the amount of over-the-counter derivatives in the last quarter of the year; Goel et al. (2022), who find that less profitable banks contracted in response to higher capital surcharges while profitable banks continued to increase their systemic importance; Dzhagityan and Orekhov (2022), who found that the GSIB surcharge – among other policy reforms – contributed to financial stability; Gündüz (2023), who finds a temporary increase in a bank’s CDS spread increase after an increase in the GSIB surcharge; and Poledna et al. (2017), who argue that changing network topology would be a better approach to improve financial fragility than the GSIB surcharge.3 Our findings are consistent with previous findings in the literature, as we find that proximity to an increase in the GSIB surcharge is associated with reductions in the systemic risk indicators of 3 Other papers have assessed the impact of the GSIB surcharge on other economic outcomes. For example, Favara et al. (2021) found that GSIBs lent less to corporates after the introduction of the surcharge, but that lending to corporates did not decrease because corporate borrowers switched to other banks. Also, Degryse et al. (2023) found that GSIBs lowered their lending to corporates after being designed as a GSIB, but argue that this effect is mostly due to stricter supervision rather than to the GSIB surcharge.

GSIBs in some cases. Still, the focus and methodology of our analysis are distinct from those of previous studies. The remainder of this note is organized as follows: Section 2 discusses the motivations behind the GSIB capital surcharge and the calculation of the GSIB score. Section 3 describes our data. Section 4 describes our empirical methodology and presents our regression results. Section 5 concludes and discusses future avenues of research. 2 – Background In the aftermath of the global financial crisis, bank capital regulations underwent significant reform to mitigate future strain on the financial system. In response to concerns regarding the systemic risk posed by the largest, most interconnected banks, global financial regulators introduced the GSIB surcharge, which requires these banks to maintain higher regulatory capital ratios than smaller, less complex banks. This surcharge requires GSIBs to internalize the costs that their failure can pose to the broader financial system, thereby reducing systemic risk by making the failure of GSIBs less likely. The GSIB surcharge also aims to be proportional to the systemic footprint of each covered bank, incentivizing these banks to reduce the systemic risk they pose. Under the Basel framework, a bank’s GSIB score is based on five equally weighted categories of systemic importance (Basel Committee on Banking Supervision 2021): (1) cross-jurisdictional activity, (2) size, (3) interconnectedness, (4) substitutability/financial institution infrastructure, and (5) complexity. These categories are further subdivided into individual indicators, each with its respective indicator weighting. Table 1. GSIB Score Systemic Indicator Weights Category Systemic Indicator Indicator Weight Size Total exposure 20% Intra-financial system assets 6.67% Interconnectedness Intra-financial system liabilities 6.67% Securities outstanding 6.67% Payments activity 6.67% Assets under custody 6.67% Substitutability Underwritten transactions in debt and equity 6.67% markets Notional amount of over-the-counter (OTC) 6.67% derivatives Complexity Trading and available-for-sale securities 6.67% Level 3 assets 6.67% Cross-jurisdictional claims 10% Cross-jurisdictional activity Cross-jurisdictional liabilities 10% Basel Committee on Banking Supervision, 2021.

To calculate each bank's indicator score, the indicator amount is divided by the aggregate global indicator amount.4 Subsequently, this amount is multiplied by 10,000 to express the indicator score in terms of basis points. 𝑏𝑏𝐼𝐼𝐼𝐼𝑏𝑏 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 𝐼𝐼𝑎𝑎𝐼𝐼𝑎𝑎𝐼𝐼𝐼𝐼 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 𝑠𝑠𝐼𝐼𝐼𝐼𝐼𝐼𝑠𝑠 = ×10,000 Then, the category scores are calcu𝐼𝐼la𝑎𝑎te𝑎𝑎d𝐼𝐼 𝑠𝑠b𝑎𝑎y𝐼𝐼 a𝐼𝐼v𝑠𝑠 e𝑎𝑎r𝑔𝑔a𝐼𝐼g𝑏𝑏in𝐼𝐼g𝑔𝑔 𝐼𝐼t𝐼𝐼h𝐼𝐼e𝐼𝐼 𝐼𝐼re𝐼𝐼s𝐼𝐼p𝐼𝐼e𝐼𝐼c 𝐼𝐼ti𝑎𝑎ve𝐼𝐼 𝑎𝑎in𝐼𝐼d𝐼𝐼icator scores for each of the five categories of systemic importance (size, interconnectedness, substitutability/financial institution infrastructure, complexity, and cross-jurisdictional activity).5 The final GSIB score is an average of five category scores: 𝐺𝐺𝐺𝐺𝐼𝐼𝐺𝐺 𝑠𝑠𝐼𝐼𝐼𝐼𝐼𝐼𝑠𝑠 𝑠𝑠𝐼𝐼𝑠𝑠𝑠𝑠+𝐼𝐼𝐼𝐼𝐼𝐼𝑠𝑠𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑠𝑠𝐼𝐼𝐼𝐼𝑠𝑠𝐼𝐼𝐼𝐼𝑠𝑠𝑠𝑠𝑠𝑠+𝑠𝑠𝑎𝑎𝑏𝑏𝑠𝑠𝐼𝐼𝐼𝐼𝐼𝐼𝑎𝑎𝐼𝐼𝐼𝐼𝑏𝑏𝐼𝐼𝑔𝑔𝐼𝐼𝐼𝐼𝑠𝑠+𝐼𝐼𝐼𝐼𝑎𝑎𝑐𝑐𝑔𝑔𝑠𝑠𝑐𝑐𝐼𝐼𝐼𝐼𝑠𝑠+𝐼𝐼𝐼𝐼𝐼𝐼𝑠𝑠𝑠𝑠 𝑗𝑗𝑎𝑎𝐼𝐼𝐼𝐼𝑠𝑠𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑔𝑔 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑎𝑎𝐼𝐼𝐼𝐼𝑠𝑠 = The GSIB framework employs a cut-off score of 130bps for GSIB designation, with each 5 subsequent bucket increment set at 100bps (Basel Committee on Banking Supervision 2013). The GSIB buckets are displayed below. Table 2. GSIB Framework Method 1 Bucketing Thresholds GSIB Buckets Bucket 5 (+3.5% CET1) 530-629 Bucket 4 (+2.5% CET1) 430-529 Bucket 3 (+2.0% CET1) 330-429 Bucket 2 (+1.5% CET1) 230-329 Bucket 1 (+1.0% CET1) 130-229 Basel Committee on Banking Supervision, 2013. The Basel GSIB surcharge standard is implemented in the United States as the “method 1” surcharge requirement for US GSIBs. US bank holding companies identified as a GSIB under the method 1 score calculation must also calculate a method 2 score that is used to set an alternative GSIB surcharge. The GSIB’s method 2 score calculation includes the same systemic risk indicators as method 1 except for the substitutability category, which is replaced by a measure of short-term wholesale funding. The individual systemic indicator scores are equal to the reported amount of the indicators multiplied by the respective coefficients presented below.6 4 The aggregate global amount for each indicator is obtained by summing the indicator amounts for largest 75 banks by asset size in the data collected by the Basel Committee. 5 The Basel Committee capped the contribution of the substitutability category to an individual bank at 100 basis points. 6 The calibration methodology employed to determine these coefficients is explained in Board of Governors of the Federal Reserve System (2015).

Table 3. US Method 2 GSIB Score Systemic Indicator Weights Category Systemic Indicator Coefficient value (%) Size Total exposures 4.423 Intra-financial system assets 12.007 Interconnectedness Intra-financial system liabilities 12.490 Securities outstanding 9.056 Notional amount of over-the-counter (OTC) 0.155 derivatives Complexity Trading and AFS securities 30.169 Level 3 assets 161.177 Cross-jurisdictional claims 9.277 Cross-jurisdictional activity Cross-jurisdictional liabilities 9.926 Code of Federal Regulations, 2016. The GSIB’s short-term wholesale funding score is calculated by dividing the average weighted short-term wholesale funding amount by the GSIB’s risk-weighted assets and then multiplying by a fixed factor of 350. Similar to method 1 GSIB scores, US GSIB scores under method 2 equal the sum of the individual systemic indicator scores. Table 4 presents the bucket thresholds for applying the method 2 surcharges. Table 4. GSIB Framework Method 2 Bucketing Thresholds GSIB Buckets Below 130 0.0% 130-229 1.0% 230-329 1.5% 330-429 2.0% 430-529 2.5% 530-629 3.0% 630-729 3.5% 730-829 4.0% 830-929 4.5% 930-1029 5.0% 1030-1129 5.5% 1130 and above 6.5% + 0.5% for every extra 100bp Code of Federal Regulations, 2017. To assess the impact of the Basel GSIB surcharge on the systemic risk of global banks, we use bank-level data from the Bank for International Settlements website, including bank systemic risk indicator values and global denominators.7 Our panel dataset includes 683 “bank-year” observations and 93 banks, covering 2013 to 2021. Of these observations, we include 276 “bankyear” observations (referring to 40 banks) in our regression analysis, as we exclude observations 7 Bank indicator values can be found here: https://www.bis.org/bcbs/gsib/gsib_assessment_samples.htm. Global denominator values can be found here: https://www.bis.org/bcbs/gsib/denominators.htm.

where the bank started with a GSIB score below 100 bps (banks with a score below 100 bps never surpassed 130 bps – the threshold score to be classified as a GSIB – in the following year). In addition, we consider method 2 data for the eight US bank holding companies that have been deemed GSIBs since the introduction of the US GSIB surcharge requirement (BNY Mellon, Bank of America, Citigroup, Goldman Sachs, J.P. Morgan, Morgan Stanley, State Street, and Wells Fargo). Our panel dataset on the annual changes in systemic risk indicators includes 40 “firm-year” observations as all eight of these bank holding companies have consistently classified as GSIBs from 2016 to 2021. Figure 1 displays how the global denominators – in the size, cross-jurisdictional activity, interconnectedness, substitutability, and complexity categories – have moved between 2013 and 2021.

Figure 1. Individual Indicator Global Denominators from 2013 to 2021

Table 5. Global Denominator Growth Rates (2013 to 2021) Category Systemic Indicator Growth Rate Size Total exposure 48.5% Intra-financial system assets 22.2% Interconnectedness Intra-financial system liabilities 25.8% Securities outstanding 42.5% Payments activity 51.2% Assets under custody 108.5% Substitutability Underwritten transactions in debt and equity 94.6% markets Notional amount of over-the-counter (OTC) -9.7% derivatives Complexity Trading and available-for-sale securities 11.9% Level 3 assets 4.8% Cross-jurisdictional claims 50.0% Cross-jurisdictional activity Cross-jurisdictional liabilities 44.4% A majority of the global denominators increased gradually from 2013 to 2021. The indicators that grew the most were assets under custody (108.5%), underwritten transactions in debt and equity markets (94.6%), and payments activity (51.2%). The notional amount of OTC derivatives decreased in absolute terms over this period (-9.7%). Overall, most indicators grew faster than the combined rate of inflation and economic growth across the world in this period, which suggests that the introduction of the GSIB surcharge did not meaningfully limit large international banks ability to grow. 4 – Regression analysis If the surcharge motivates GSIBs to reduce their systemic risk indicators, firms whose GSIB score is closer to triggering an increase or decrease in the surcharge would likely be more motivated to manage their activities to achieve a reduction in the surcharge. Therefore, a negative association between proximity to surcharge thresholds and growth in systemic risk indicators would be expected. To test whether this association happens in practice, we perform regressions that estimate how the growth rate of firms’ systemic risk indicators relates to how close the firm’s GSIB score is to triggering an increase or decrease in the surcharge. Specifically, we estimate the following regression equation: In this equ∆a%ti𝑅𝑅on𝐼𝐼𝑠𝑠,𝑏𝑏 t𝐼𝐼h𝐼𝐼e𝐼𝐼 d𝐼𝐼𝐼𝐼e𝐼𝐼p𝐼𝐼e𝐼𝐼n𝐼𝐼d𝑖𝑖𝑖𝑖en=t v𝛽𝛽a0ri+ab𝛽𝛽l1e𝐿𝐿 i𝐼𝐼s𝐿𝐿 th𝐺𝐺e𝐼𝐼 𝐼𝐼p𝐼𝐼e𝑠𝑠r𝑖𝑖c,𝑖𝑖e−n1ta+g𝛽𝛽e 2c𝐻𝐻h𝐼𝐼a𝑎𝑎nℎg𝐺𝐺e𝐼𝐼 i𝐼𝐼n𝐼𝐼 𝑠𝑠o𝑖𝑖n,𝑖𝑖e− 1o+f th𝛾𝛾𝑖𝑖e +sy𝛼𝛼s𝑖𝑖te+m i𝜖𝜖c𝑖𝑖 𝑖𝑖risk indicators for bank i between t-1 and t, denoted as RiskIndicator .8 The main explanatory i,t variables are indicator variables for (1) whether the t-1 GSIB score of a firm was close to the Δ% threshold at which the surcharge would decrease (LowScore ) and (2) whether the t-1 GSIB i,t-1 score of a firm was close to the threshold at which the surcharge would increase (HighScore ). i,t-1 LowScore is calculated as follows: i,t-1 8 Δ%RiskIndicator = (RiskIndicator - RiskIndicator ) / RiskIndicator . i,t i,t i,t-1 i,t-1

𝐿𝐿𝐼𝐼𝐿𝐿𝐺𝐺𝐼𝐼𝐼𝐼𝐼𝐼𝑠𝑠𝑖𝑖,𝑖𝑖−1 (𝑏𝑏𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑎𝑎 𝐼𝐼𝑖𝑖 𝐼𝐼𝐼𝐼𝑠𝑠 𝑠𝑠𝐼𝐼𝐼𝐼𝐼𝐼𝑠𝑠 𝑏𝑏𝑎𝑎𝐼𝐼𝑏𝑏𝑠𝑠𝐼𝐼 𝐼𝐼𝐼𝐼𝐼𝐼𝑎𝑎𝑠𝑠−1) 1,𝐼𝐼𝑖𝑖 ≥𝐺𝐺𝐺𝐺𝐼𝐼𝐺𝐺𝐺𝐺𝐼𝐼𝐼𝐼𝐼𝐼𝑠𝑠𝑖𝑖,𝑖𝑖−1 ≥𝑏𝑏𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑎𝑎 𝐼𝐼𝑖𝑖 𝐼𝐼𝐼𝐼𝑠𝑠 𝑠𝑠𝐼𝐼𝐼𝐼𝐼𝐼𝑠𝑠 𝑏𝑏𝑎𝑎𝐼𝐼𝑏𝑏𝑠𝑠𝐼𝐼 𝐼𝐼𝐼𝐼𝐼𝐼𝑎𝑎𝑠𝑠 =� 0.95 For example, for the first GSIB score bucket 0(,1𝐼𝐼3𝐼𝐼ℎ0𝑠𝑠 t𝐼𝐼o𝐿𝐿 2𝐼𝐼𝑠𝑠2𝑠𝑠9), GSIB scores between 130 and 135.8 (which equals 129/0.95) are considered close to the low threshold and scores above 135.8 are not.9 Conversely, HighScore is calculated as follows: i,t-1 𝐻𝐻𝐼𝐼𝑎𝑎ℎ𝐺𝐺𝐼𝐼𝐼𝐼𝐼𝐼𝑠𝑠𝑖𝑖,𝑖𝑖−1 1,𝐼𝐼𝑖𝑖 𝐼𝐼𝐼𝐼𝑐𝑐 𝐼𝐼𝑖𝑖 𝐼𝐼𝐼𝐼𝑠𝑠 𝑠𝑠𝐼𝐼𝐼𝐼𝐼𝐼𝑠𝑠 𝑏𝑏𝑎𝑎𝐼𝐼𝑏𝑏𝑠𝑠𝐼𝐼 𝐼𝐼𝐼𝐼𝐼𝐼𝑎𝑎𝑠𝑠 ≥𝐺𝐺𝐺𝐺𝐼𝐼𝐺𝐺𝐺𝐺𝐼𝐼𝐼𝐼𝐼𝐼𝑠𝑠𝑖𝑖,𝑖𝑖−1 ≥(𝐼𝐼𝐼𝐼𝑐𝑐 𝐼𝐼𝑖𝑖 𝐼𝐼𝐼𝐼𝑠𝑠 𝑠𝑠𝐼𝐼𝐼𝐼𝐼𝐼𝑠𝑠 𝑏𝑏𝑎𝑎𝐼𝐼𝑏𝑏𝑠𝑠𝐼𝐼 𝐼𝐼𝐼𝐼𝐼𝐼𝑎𝑎𝑠𝑠+1)/1.05 =� To account for time-specific differences in th0e, 𝐼𝐼a𝐼𝐼vℎe𝑠𝑠r𝐼𝐼a𝐿𝐿g𝐼𝐼e𝑠𝑠 𝑠𝑠growth of systemic risk indicators, we introduce year fixed effects into our regression model, denoted by γ above. In addition, we t include bank-specific fixed effects in some of our specifications, denoted by α above. These i bank-specific fixed effects control for unobserved, time-invariant firm-specific characteristics that may influence the dependent variable. For this regression analysis, we exclude observations where a firm’s GSIB score in year t-1 is below 100 bps. GSIB designation has a cut-off score of 130 bps, and a firm with a score below 100 bps never surpassed 130 bps in the following year. Therefore, we exclude these observations because these firms’ activity level is unlikely to be directly affected by the GSIB surcharge requirement. In addition, we exclude observations where the average share of the systemic risk indicator in the firm’s overall GSIB score is below 1%.10 This exclusion ensures that regression estimates are not driven by GSIBs with very little activity relating to the specific systemic risk indicator. Regressions based on all GSIBs worldwide Table 6 presents our main regression results.11 In addition to specification 1, which includes firm fixed effects, we also present results based on specification 2 which omits them. This omission implies that the effects of the proximity variables are informed by variation across GSIBs in this specification; however, other idiosyncratic, constant differences across firms are not controlled for. Meanwhile, we include the relevant systemic risk indicator’s share of the GSIB score as a 9 We have chosen to set low and high score indicator variables based on a percentage distance from the bucket thresholds rather than based on absolute distance because the same absolute change in score (e.g., ten score points) represents a much bigger relative change for a firm with a low GSIB score than for a firm with a high GSIB score. 10 To determine the average share of a systemic risk indicator in a firm’s overall GSIB score we start by calculating, for each year, the ratio between the score contribution of a systemic risk indicator and a firm’s full GSIB score. Then, we average this ratio over the years in the analysis. 11 As an alternative to defining the dependent variable in our regressions as the percentage change in a systemic risk indicator, we defined the dependent variable as the change of systemic risk indicator’s share of a firm’s GSIB score in regressions presented in Appendix B. Regression results are qualitatively similar.

control variable in specification 2. The existing share of a systemic risk indicator may influence its future path, either because it is a predictor of further expansion or of reduction in exposure. Table 6. Effect of Proximity to Surcharge Increase or Decrease on Systemic Risk Indicators Explanatory Variables (1) (2) Dependent Variables Indicator N LowScore i,t-1 HighScore i,t-1 LowScore i,t-1 HighScore i,t-1 Share of GSIB score 0.035** 0.009 0.039*** 0.012 0.289*** Δ% total exposuresi,t 276 (0.013) (0.015) (0.012) (0.012) (0.054) Δ% intra-financial system 0.109 -0.041 0.052 -0.077*** -2.174*** 276 assetsi,t (0.071) (0.038) (0.059) (0.027) (0.629) Δ% intra-financial system -0.020 -0.021 -0.015 -0.058 -0.784** 276 liabilitiesi,t (0.045) (0.040) (0.042) (0.036) (0.377) 0.053* 0.001 0.039 -0.005 0.221 Δ% securities outstandingi,t 276 (0.029) (0.022) (0.027) (0.020) (0.404) -0.015 -0.019 0.000 -0.022 -0.319 Δ% payments activityi,t 276 (0.034) (0.034) (0.032) (0.031) (0.208) 1.174 -0.294 0.950 -0.076 -0.387 Δ% assets under custodyi,t 234 (1.134) (0.226) (0.966) (0.091) (0.310) Δ% underwritten transactions in -0.080 -0.112 -0.038 -0.077* -2.771* 256 debt and equity marketsi,t (0.063) (0.068) (0.062) (0.043) (1.401) Δ% notional amount of OTC 0.043 0.016 0.039 -0.006 -1.046*** 219 derivativesi,t (0.030) (0.028) (0.031) (0.022) (0.309) -0.083 -0.187** -0.013 -0.052 -4.250*** Δ% trading and AFS securitiesi,t 276 (0.128) (0.088) (0.090) (0.077) (1.477) -0.169 0.310 -0.047 0.443 -3.839** Δ% level 3 assetsi,t 268 (0.126) (0.440) (0.083) (0.513) (1.866) -0.003 0.022 0.012 0.005 -0.532*** Δ% cross-jurisdictional claimsi,t 271 (0.037) (0.028) (0.029) (0.026) (0.164) Δ% cross-jurisdictional -0.020 -0.004 -0.005 0.020 -0.652*** 276 liabilitiesi,t (0.049) (0.037) (0.045) (0.030) (0.213) Year Fixed Effects Yes Yes Firm Fixed Effects Yes No Notes: Standard errors in parenthesis. * = p-value < 0.1; ** = p-value < 0.05; *** = p-value < 0.01. Appendix A presents the descriptive statistics of the variables used in the regression analyses of this paper. In specification 1 (with firm fixed effects), HighScore is a statistically significant predictor of i,t-1 the growth of a GSIB’s trading and AFS securities (5% significance). Proximity to an increase in the GSIB surcharge is associated with an 18.7 percentage points (pp) lower growth rate of a firm’s trading and AFS securities. Meanwhile, LowScore is a statistically significant predictor i,t-1 of the growth of a GSIB’s total exposures (5% significance) and securities outstanding (10% significance). Proximity to a decrease in the GSIB surcharge is associated with a 3.5 pp higher growth rate of a firm’s total exposures and with a 5.3 pp higher growth rate of a firm’s securities outstanding.

In specification 2 (without firm fixed effects), HighScore is a statistically significant predictor i,t-1 of the growth of a GSIB’s intra-financial system assets (1% significance) and underwritten transactions in debt and equity (10% significance). Proximity to an increase in the GSIB surcharge is associated with a 7.7 pp lower growth rate of a firm’s intra-financial system assets and with a 7.7 pp lower growth rate of a firm’s underwritten transactions in debt and equity. Consistent with specification 1, LowScore is also a statistically significant predictor of the i,t-1 growth of a GSIB’s total exposures (1% significance) in specification 2. Proximity to a decrease in the GSIB surcharge is associated with a 3.9 pp higher growth rate of a firm’s total exposures. Specification 2 also shows that firms tend to reduce the systemic risk indicators that comprise a high share of their GSIB score. The score shares of intra-financial system assets, intra-financial system liabilities, underwritten transactions in debt and equity, OTC derivatives, trading and AFS securities, level 3 assets, cross-jurisdictional claims, and cross-jurisdictional liabilities are all negatively associated with the ensuing growth rate of these systemic risk indicators. Total exposures are an exception. Higher score shares of total exposures are associated with higher growth rate in total exposures in the ensuing year. To understand whether proximity to an increase or decrease in the surcharge affects how firms manage their GSIB surcharge as a whole, we regress a modified version of the GSIB score on the proximity variables. In calculating the GSIB scores used as the dependent variable, we fix the global denominators at their 2013 values. This modification to the GSIB score calculation neutralizes the effect of changes in global denominators on scores, thereby ensuring that the year-on-year measured change in a firm’s GSIB score is due to the changes in its systemic risk indicators. Note that we do not modify the GSIB scores used in calculating the proximity variables (which, therefore, continue to reflect the proximity of a firm’s GSIB score at a particular point in time to the surcharge bucket thresholds). Table 7 presents the results of this regression analysis, including a specification that includes firm fixed effects (specification 1) and an alternative specification that omits firm fixed effects (specification 2). Table 7. Effect of Proximity to Surcharge Increase or Decrease on Firm-Level GSIB Score Explanatory Variables Dependent Variables (1) (2) LowScore HighScore LowScore HighScore i,t-1 i,t-1 i,t-1 i,t-1 0.026 -0.009 0.027 -0.009 Δ% GSIB scorei,t (0.021) (0.017) (0.016) (0.016) Year Fixed Effects Yes Yes Firm Fixed Effects Yes No Notes: N = 276 in all regressions. Standard errors in parenthesis. * = p-value < 0.1; ** = p-value < 0.05; *** = pvalue < 0.01. Appendix A presents the descriptive statistics of the variables used in the regression analyses of this paper. Similar to most of the regressions for individual systemic risk indicators, the regressions with growth in the GSIB score as the dependent variable do not show a statistically significant association between the dependent variable and the proximity variables. Taken together, the

regression results in Tables 6 and 7 suggest that large banks worldwide generally have not meaningfully changed the growth of their activities relating to systemic risk indicators to manage their GSIB surcharge. The main exceptions are possible reductions on the growth of intrafinancial system assets and trading and ASF securities for firms close to an increase in their GSIB surcharge and increases in the growth of total exposures for firms in the lower score range of a GSIB surcharge bucket. The association of total exposures with firms in the lower range of a GSIB surcharge bucket may result from a firm’s decision to cross a GSIB surcharge bucket (and, therefore, temporarily score in the lower end of a GSIB surcharge bucket) being associated with a plan for sustained growth of a GSIB’s activities. To further assess the effect of the GSIB surcharge on the systemic risk posed by firms, we consider whether proximity to a change in the GSIB surcharge affects a firm’s SRISK (which represents a firm’s expected capital shortfall in a potential future financial crisis).12 If a firm constrains its systemic risk to avoid an increase in the GSIB surcharge, then a negative association between proximity to a change in the GSIB surcharge and SRISK may be expected. Table 8 presents the results of this regression analysis, including a specification that includes firm fixed effects (specification 1) and an alternative specification that omits firm fixed effects (specification 2). Table 8. Effect of Proximity to Surcharge Increase or Decrease on a Firm’s SRISK Explanatory Variables Dependent Variables (1) (2) LowScore HighScore LowScore HighScore i,t-1 i,t-1 i,t-1 i,t-1 -0.232 0.040 -0.281 -0.142 Δ% SRISKi,t (0.233) (0.321) (0.282) (0.332) Year Fixed Effects Yes Yes Firm Fixed Effects Yes No Notes: N = 264 in all regressions. Standard errors in parenthesis. * = p-value < 0.1; ** = p-value < 0.05; *** = pvalue < 0.01. Appendix A presents the descriptive statistics of the variables used in the regression analyses of this paper. The regressions with the percentage change in SRISK as the dependent variable do not indicate a statistically significant association between the dependent variable and the proximity variables. This suggests that GSIB surcharge is not providing strong incentives for firms to constrain their systemic risk, as measured by SRISK. 12 SRISK data comes from NYU Stern’s V-Lab Systemic Risk Analysis: https://vlab.stern.nyu.edu/srisk. Note that SRISK data is not available for some of the GSIBs. See Brownlees and Engle (2017) for a description of the SRISK framework.

Regressions based on the GSIB method 2 score applied to US GSIBs To complement the analysis based on the Basel GSIB surcharge requirement, we perform a similar regression analysis based on the method 2 indicators of the eight US GSIBs.13 Table 9 displays the results. These regressions use data from 2016 to 2021.14 Table 9. Effect of Proximity to Surcharge Increase or Decrease on Method 2 Systemic Risk Indicators Explanatory Variables (1) (2) Dependent Variables Indicator N LowScore HighScore LowScore HighScore Share of i,t-1 i,t-1 i,t-1 i,t-1 GSIB score -0.015 -0.002 -0.011 -0.016 -0.115 Δ% total exposuresi,t 40 (0.029) (0.020) (0.018) (0.016) (0.111) Δ% intra-financial system -0.127 -0.213 -0.074 -0.129 -3.456 40 assetsi,t (0.119) (0.141) (0.103) (0.118) (1.865) Δ% intra-financial system 0.021 0.027 0.014 0.007 0.307 35 liabilitiesi,t (0.069) (0.067) (0.064) (0.055) (0.747) 0.050 0.203** 0.067 0.103* -0.231 Δ% securities outstandingi,t 40 (0.042) (0.080) (0.046) (0.053) (0.919) Δ% amount of OTC 0.024 0.170 0.012 0.120 -1.309*** 35 derivativesi,t (0.042) (0.094) (0.032) (0.070) (0.351) Δ% trading and AFS -0.079 -0.100 -0.112 -0.152* -1.059 40 securitiesi,t (0.124) (0.073) (0.092) (0.067) (0.840) 0.0173 0.002 -0.031 -0.016 -1.260 Δ% level 3 assetsi,t 30 (0.224) (0.201) (0.148) (0.157) (1.020) 0.047 -0.044 -0.003 -0.060 0.165 Δ% cross-jurisdictional claimsi,t 40 (0.052) (0.058) (0.044) (0.047) (0.227) Δ% cross-jurisdictional 0.027 -0.086 0.002 -0.136 0.469 40 liabilitiesi,t (0.105 (0.095) (0.089) (0.078) (0.724) Δ% short-term wholesale -0.065 -0.052 -0.054* -0.061 -0.094 40 fundingi,t (0.037) (0.047) (0.026) (0.033) (0.065) -0.032 -0.051 -0.034 -0.063 Δ% method 2 GSIB scorei,t --- 40 (0.056) (0.048) (0.039) (0.034) Year Fixed Effects Yes Yes Firm Fixed Effects Yes No Notes: Standard errors in parenthesis. * = p-value < 0.1; ** = p-value < 0.05; *** = p-value < 0.01. Appendix A presents the descriptive statistics of the variables used in the regression analyses of this paper. In specification 1 (with firm fixed effects), HighScore is a statistically significant predictor of i,t-1 the growth of a GSIB’s securities outstanding (5% significance). Proximity to an increase in the GSIB surcharge is associated with a 20.3 pp higher growth rate of a firm’s securities outstanding. 13 The method 2 sample includes the following US GSIBs: BNY Mellon, Bank of America, Citigroup, Goldman Sachs, J.P. Morgan, Morgan Stanley, State Street, and Wells Fargo. 14 Note that, like in the regressions of Table 6, observations are excluded when a systemic risk indicator does not contribute at least 1% to a firm’s GSIB score on average. Therefore, the observations of a few firms are excluded from some of the regressions.

Meanwhile, LowScore is not a statistically significant predictor of any of the method 2 i,t-1 systemic risk indicators for US GSIBs. In specification 2 (without firm fixed effects), HighScore is a statistically significant predictor i,t-1 of the growth of a GSIB’s securities outstanding (10% significance) and trading and AFS securities (10% significance). Proximity to an increase in the GSIB surcharge is associated with a 10.3 pp higher growth rate of a firm’s securities outstanding and with a 15.2 pp lower growth rate of a firm’s trading and AFS securities. Meanwhile, LowScore is a statistically significant i,t-1 predictor of the growth of a GSIB’s short-term wholesale funding score (10% significance). Proximity to a decrease in the GSIB surcharge is associated with a 5.4 pp lower growth rate of a firm’s short-term wholesale funding score. The regressions based on US GSIB surcharge method 2 systemic risk indicators generally do not provide much evidence of an effect of proximity to changes in the surcharge to changes on systemic risk indicators. The lack of statistical significance of effects may partly be due to lack of power of the regressions, as we only have 40 observations at most per regression. 5 – Discussion Our analysis provides limited evidence that the GSIB surcharge has affected global banks’ management of their exposures. For the worldwide sample of GSIBs, we find some association between proximity to an increase in surcharge and a decrease in the growth of intra-financial system assets, underwriting activities, and trading and AFS securities. In the case of US GSIBs and the method 2 GSIB surcharge, we find some association between proximity to an increase in surcharge and a decrease in the growth of trading and AFS securities as well as some association between proximity to the lower bound of a score bucket and a decrease in short-term wholesale funding. We also find a positive association between proximity to lower bound of the surcharge bucket and (1) increase in total exposures and securities outstanding for the worldwide sample of GSIBs and (2) increase in securities outstanding for US GSIBs. Overall, our regression results should be interpreted cautiously. Some of the effects are not consistent across specifications, and some are only statistically significant at 10% significance. This study aims to enhance the understanding of how the GSIB surcharge affects the systemic risk posed by the largest, most complex international banks. In attempting to do so, we focus on how the GSIB surcharge affects firms’ systemic risk indicators over time. This analysis helps understand how banks adjust their activities in response to the capital incentives introduced by the surcharge. Still, this analysis only helps answer how the surcharge affects bank systemic risk to the extent that the indicators considered appropriately measure such risk. We test whether proximity to a GSIB surcharge change affects how a firm’s SRISK changes over time and find no effect. There have been numerous other attempts in the academic literature to estimate the systemic risk posed by individual banks, which often have significant limitations (see a discussion in Hawley and Migueis 2021). To the extent that better measures of the systemic risk posed by individual banks are available, future research should assess how the GSIB surcharge affects them.

References Basel Committee on Banking Supervision (2013). GSIB Framework: Cut-off score and bucket thresholds. https://www.bis.org/bcbs/gsib/cutoff.htm. Basel Committee on Banking Supervision (2021). SCO Scope and definitions SCO40 Global Systemically Important Banks. https://www.bis.org/basel_framework/chapter/SCO/40.htm?inforce=20211109&published=2021 1109&tldate=20250920. Behn, Markus, and Alexander Schramm (2021). The impact of G-SIB identification on bank lending: Evidence from syndicated loans. Journal of Financial Stability 57, 100930. https://doi.org/10.1016/j.jfs.2021.100930. Behn, Markus, Giacomo Mangiante, Laura Parisi, and Michael Wedow (2021). Behind the Scenes of the Beauty Contest – Window Dressing and the G-SIB Framework. International Journal of Central Banking Vol. 18 No. 5, pages 301-342. https://www.ijcb.org/journal/ijcb22q5a7.htm. Berry, Jared, Akber Khan, and Marcelo Rezende (2024). How Do Global Systemically Important Banks Lower Capital Surcharges? Journal of Financial Services Research. https://doi.org/10.1007/s10693-024-00426-w. Board of Governors of the Federal Reserve System (2015). Calibrating the G-SIB Surcharge. https://www.federalreserve.gov/aboutthefed/boardmeetings/gsib-methodology-paper- 20150720.pdf. Brogi, M., Lagasio, V., Porretta, P., & Riccetti, L. (2017). Systemic Risk Measurement: Bucketing GSIBs between literature and supervisory view. Social Science Research Network. https://doi.org/10.2139/ssrn.2915172. Brownlees, Christian, and Robert Engle (2017). SRISK: A Conditional Capital Shortfall Measure of Systemic Risk. The Review of Financial Studies Vol. 30 No. 1, pages 48-79. https://doi.org/10.1093/rfs/hhw060. Code of Federal Regulations, 2016. § 217.405 Method 2 score. Available at: eCFR :: 12 CFR 217.404 -- Method 1 score. Code of Federal Regulations, 2016. § 217.405 Method 2 score. Available at: eCFR :: 12 CFR 217.405 -- Method 2 score. Code of Federal Regulations, 2017. § 217.403 GSIB surcharge. Available at: https://www.ecfr.gov/current/title-12/chapter-II/subchapter-A/part-217/subpart-H/section- 217.403. Degryse, Hans, Mike Mariathasan, and Hien T. Tang (2023). GSIB status and corporate lending. Journal of Corporate Finance 80, 102362. https://doi.org/10.1016/j.jcorpfin.2023.102362.

Dzhagityan, E., & Orekhov, M. (2022). Global Systemically Important Banks: Do they still pose risks for financial stability? International Organisations Research Journal, 17(3), 48– 74. https://doi.org/10.17323/1996-7845-2022-03-03. Favara, G., Ivanov, I. T., & Rezende, M. (2021). GSIB surcharges and bank lending: Evidence from US corporate loan data. Journal of Financial Economics, 142(3), 1426– 1443. https://doi.org/10.1016/j.jfineco.2021.06.026. Garcia, Luis, Ulf Lewrick, and Taja Secnik (2023). Window Dressing and the Designation of Global Systemically Important Banks. Journal of Financial Services Research 64, pages 231- 264. https://doi.org/10.1007/s10693-023-00417-3. Goel, Tirupam, Ulf Lewrick, and Aakriti Mathur (2019). Playing it safe: global systemically important banks after the crisis. BIS Quarterly Review, September 2019. https://www.bis.org/publ/qtrpdf/r_qt1909e.htm. Goel, Tirupam, Ulf Lewrick, and Aakriti Mathur (2022). Does regulation only bite the less profitable? Evidence from the too-big-to-fail reforms. Bank of England, Staff Working Paper No. 946. https://www.bankofengland.co.uk/-/media/boe/files/working-paper/2021/doesregulation-only-bite-the-less-profitable-evidence-from-the-too-big-to-fail-reforms.pdf. Gündüz, Y. (2020). The market impact of systemic risk capital surcharges. European Financial Management 29, pages 1401-1440. DOI: 10.1111/eufm.12398. Hawley, Andrew, and Marco Migueis (2021). Measuring the systemic importance of large US banks. Board of Governors of the Federal Reserve System, FEDS Notes. https://doi.org/10.17016/2380-7172.2988. Ho, Kelvin, Eric Wong, and Edward Tan (2022). Complexity of global banks and the implications for bank risk: Evidence from foreign banks in Hong Kong. Journal of Banking and Finance 134, 106034. https://doi.org/10.1016/j.jbankfin.2020.106034. Poledna, S., Bochmann, O., & Thurner, S. (2017). Basel III capital surcharges for GSIBs are far less effective in managing systemic risk in comparison to network-based, systemic riskdependent financial transaction taxes. Journal of Economic Dynamics and Control 77, pages 230-246. DOI: 10.1016/j.jedc.2017.02.004. Violon, Aurelien, Dominique Durant, and Oana Toader (2020). International Journal of Central Banking Vol. 16 No. 5, pages 95-142. https://www.ijcb.org/journal/ijcb20q4a3.htm.

Appendix A Table A.1 – Summary Statistics for Method 1 Regression Variables Summary Statistics Dependent Variables N Mean SD Min Max P25 P50 P75 Δ% total exposures 276 0.0387 0.0973 -0.3121 0.3468 -0.0157 0.0442 0.0920 Δ% intra-financial system assets 276 0.0443 0.2392 -0.5293 1.3639 -0.0922 0.0272 0.1387 Δ% intra-financial system liabilities 276 0.0379 0.2184 -0.5045 1.0196 -0.0938 0.0110 0.1181 Δ% securities outstanding 276 0.0518 0.1629 -0.5320 0.6180 -0.0413 0.0344 0.1458 Δ% payments activity 276 0.0823 0.1893 -0.4152 1.1438 -0.0204 0.0667 0.1457 Δ% assets under custody 276 0.2074 1.4967 -0.7912 22.8882 -0.0056 0.0697 0.1805 Δ% underwritten transactions in 268 0.1466 0.5330 -0.7297 6.6452 -0.0407 0.0634 0.2060 debt and equity markets Δ% notional amount of OTC 276 0.0599 0.2613 -0.4333 2.7491 -0.0750 0.0292 0.1468 derivatives Δ% trading and AFS securities 276 0.0989 0.6305 -0.8549 7.0775 -0.1535 0.0023 0.2021 Δ% level 3 assets 270 0.1020 1.2036 -1 18.484 -0.1680 -0.0221 0.1628 Δ% cross-jurisdictional claims 276 0.0808 0.1596 -0.3609 1.0043 -0.0052 0.0602 0.1383 Δ% cross-jurisdictional liabilities 276 0.0837 0.2242 -0.4663 1.2937 -0.0331 0.0556 0.1745 Δ% GSIB score 276 0.0412 0.1022 -0.2857 0.3864 -0.0211 0.0297 0.0981 Δ% SRISK 264 0.1909 1.7775 -4.5135 24.4739 -0.1471 0.0291 0.2813 low score 276 0.0978 0.2976 0 1 0 0 0 high score 276 0.1304 0.3374 0 1 0 0 0 total exposures share of GSIB score 276 0.1934 0.0858 0.0379 0.4955 0.1344 0.1767 0.2407 intra-financial system assets share 276 0.0671 0.0246 0.0129 0.1400 0.0512 0.0632 0.0794 of GSIB score intra-financial system liabilities 276 0.0668 0.0282 0.0121 0.1686 0.0490 0.0642 0.0809 share of GSIB score securities outstanding share of 276 0.0593 0.0232 0.0078 0.1325 0.0447 0.0580 0.0727 GSIB score payments activity share of GSIB 276 0.0683 0.0567 0.0096 0.3727 0.0399 0.0532 0.0769 score assets under custody share of GSIB 276 0.0801 0.1575 0.0009 0.7710 0.0137 0.0322 0.0683 score underwritten transactions in debt and equity markets share of GSIB 276 0.0696 0.0408 0 0.2007 0.0418 0.0615 0.1015 score notional amount of OTC derivatives 276 0.0717 0.0592 0.0008 0.2631 0.0204 0.0594 0.1111 share of GSIB score trading and AFS securities share of 276 0.0687 0.0393 0.0071 0.2471 0.0421 0.0657 0.0882 GSIB score level 3 assets share of GSIB score 276 0.0682 0.0490 0 0.2778 0.0336 0.0547 0.1024 cross-jurisdictional claims share of 276 0.1075 0.0614 0.0028 0.2616 0.0604 0.1104 0.1417 GSIB score cross-jurisdictional liabilities share 276 0.1060 0.0586 0.0056 0.2551 0.0578 0.0972 0.1379 of GSIB score

Table A.2 – Summary Statistics for Method 2 Regression Variables Summary Statistics Dependent Variables N Mean SD Min Max P25 P50 P75 Δ% total exposures 40 0.0453 0.1072 -0.1673 0.3468 0.0046 0.0582 0.0994 Δ% intra-financial system assets 40 0.0296 0.2004 -0.2574 0.7720 -0.1174 0.0083 0.0921 Δ% intra-financial system liabilities 40 0.0269 0.1659 -0.4576 0.3376 -0.0552 0.0189 0.1311 Δ% securities outstanding 40 0.0120 0.1704 -0.4367 0.2949 -0.0801 0.0069 0.0943 Δ% amount of OTC derivatives 40 0.0347 0.1768 -0.3157 0.5871 -0.0827 0.0229 0.1152 Δ% trading and AFS securities 40 -0.0244 0.2128 -0.5782 0.3936 -0.1584 -0.0340 0.0980 Δ% level 3 assets 35 -0.0460 0.3950 -1 1.1818 -0.2153 -0.0309 0.0661 Δ% cross-jurisdictional claims 40 0.0693 0.1362 -0.1210 0.6053 -0.0084 0.0495 0.1268 Δ% cross-jurisdictional liabilities 40 0.0697 0.2481 -0.3602 1.1337 -0.0498 0.0075 0.1404 Δ% short-term wholesale funding 40 0.0263 0.0971 -0.1684 0.3001 -0.0254 0.0097 0.0781 Δ% method 2 GSIB score 40 0.0191 0.1032 -0.2036 0.2867 -0.0553 0.0225 0.0540 low score 40 0.3 0.4641 0 1 0 0 1 high score 40 0.2 0.4051 0 1 0 0 1 total exposures share of GSIB score 40 0.1509 0.0843 0.0409 0.3057 0.0651 0.1464 0.2268 intra-financial system assets share 40 0.0430 0.0140 0.0146 0.0727 0.0330 0.0434 0.0532 of GSIB score intra-financial system liabilities 40 0.0556 0.0336 0.0080 0.1266 0.0259 0.0558 0.0843 share of GSIB score securities outstanding share of 40 0.0693 0.0355 0.0134 0.1388 0.0370 0.0689 0.0930 GSIB score amount of OTC derivatives share of 40 0.0632 0.0382 0.0038 0.1090 0.0202 0.0786 0.0965 GSIB score trading and AFS securities share of 40 0.0708 0.0325 0.0187 0.1303 0.0488 0.0710 0.0982 GSIB score level 3 assets share of GSIB score 40 0.0422 0.0320 0 0.1103 0.0211 0.0385 0.0562 cross-jurisdictional claims share of 40 0.0664 0.0338 0.0276 0.1444 0.0371 0.0610 0.0808 GSIB score cross-jurisdictional liabilities share 40 0.0672 0.0340 0.0218 0.1569 0.0487 0.0554 0.0748 of GSIB score short-term wholesale funding share 40 0.3715 0.2129 0.1432 0.6970 0.1820 0.2996 0.5874 of GSIB score

Appendix B As an alternative to the main analysis in this paper, we considered setting the dependent variable as the change in a systemic risk indicator’s share of a firm’s GSIB score. Table B.1 presents the results of regressions with this alternative dependent variable. Which of the approaches to define the dependent variable is preferable is not obvious to us and their empirical results are qualitatively similar, so we have chosen to define the dependent variable as the percentage change in the systemic risk indicator in the main analysis as that approach is a bit simpler. Table B.1 – Effect of Proximity to Surcharge Increase or Decrease on Systemic Risk Indicators and Aggregate Method 1 GSIB Score Low Score-1 and High Scoreit-1 (1) (2) Dependent Variables Indicator Share N LowScore HighScore LowScore HighScore i,t-1 i,t-1 i,t-1 i,t-1 of GSIB score N Δ total exposures score contribution 0.005** 0.001 0.005** 0.001 0.071*** 276 / GSIB scorei,t (0.002) (0.003) (0.002) (0.002) (0.013) Δ intra-financial system assets score 0.007* -0.002 0.004 -0.005** -0.147** 276 contribution / GSIB scorei,t (0.004) (0.003) (0.003) (0.002) (0.057) Δ intra-financial system liabilities 0.000 -0.000 0.000 -0.003 -0.068** 276 score contribution / GSIB scorei,t (0.002) (0.002) (0.002) (0.002) (0.029) Δ securities outstanding score 0.003 -0.000 0.002 -0.001 0.017 276 contribution / GSIB scorei,t (0.002) (0.002) (0.002) (0.002) (0.023) Δ payments activity score -0.002 0.002 -0.001 0.000 -0.014 276 contribution / GSIB scorei,t (0.003) (0.003) (0.003) (0.003) (0.018) Δ assets under custody score 0.003 -0.001 0.003* -0.000 -0.011*** 234 contribution / GSIB scorei,t (0.002) (0.001) (0.002) (0.001) (0.002) Δ underwritten transactions in debt -0.126** 256 -0.002 -0.003 -0.001 -0.002 and equity markets score (0.036) (0.004) (0.003) (0.004) (0.002) contribution / GSIB scorei,t Δ notional amount of OTC -0.070* 219 0.003 -0.000 0.002 -0.001 derivatives score contribution / (0.039) (0.003) (0.003) (0.003) (0.002) GSIB scorei,t Δ trading and AFS securities score -0.003 -0.008* 0.000 -0.004 -0.166** 276 contribution / GSIB scorei,t (0.006) (0.005) (0.004) (0.004) (0.049) Δ level 3 assets score contribution / -0.000 -0.006 0.001 -0.001 -0.163** 268 GSIB scorei,t (0.007) (0.008) (0.006) (0.008) (0.052) Δ cross-jurisdictional claims score 0.003 0.001 0.003 0.000 -0.015 271 contribution / GSIB scorei,t (0.002) (0.002) (0.002) (0.002) (0.009) Δ cross-jurisdictional liabilities 0.002 0.002 0.003 0.002 -0.032* 276 score contribution / GSIB scorei,t (0.004) (0.003) (0.004) (0.002) (0.018) 0.029 -0.010 0.030* -0.009 -- 276 Δ GSIB score / GSIB scorei,t (0.022) (0.018) (0.018) (0.018) 11.964 24.694 13.091 21.668 -- 264 Δ SRISK / GSIB scorei,t (12.801) (15.715) (13.714) (14.301) Year Fixed Effects Yes Yes -- Firm Fixed Effects Yes No -- Notes: Standard errors in parenthesis. * = p-value < 0.1; ** = p-value < 0.05; *** = p-value < 0.01.

Table B.2 – Summary Statistics for Table B1 Regression Variables Summary Statistics Dependent Variables N Mean SD Min Max P25 P50 P75 Δ total exposures score contribution 276 -0.0005 0.0151 -0.0543 0.0584 -0.0075 -0.0006 0.0074 / GSIB score Δ intra-financial system assets score 276 -0.0001 0.0142 -0.0696 0.0628 -0.0061 0.0000 .0067064 contribution / GSIB score Δ intra-financial system liabilities 276 -0.0000 0.0129 -0.0457 0.0626 -0.0064 -0.0004 0.0046 score contribution / GSIB score Δ securities outstanding score 276 0.0003 0.0071 -0.0261 0.0299 -0.0044 -0.0001 0.0037 contribution / GSIB score Δ payments activity score 276 0.0008 0.0119 -0.0664 0.0857 -0.0035 0.0005 0.0050 contribution / GSIB score Δ assets under custody score 276 -0.0006 0.0080 -0.0446 0.0327 -0.0022 -0.0002 0.0016 contribution / GSIB score Δ underwritten transactions in debt and equity markets score 268 0.0000 0.0178 -0.0842 0.1346 -0.0062 -0.0003 0.0051 contribution / GSIB score Δ notional amount of OTC derivatives score contribution / 276 0.0002 0.0130 -0.0731 0.0597 -0.0023 0.0007 0.0051 GSIB score Δ trading and AFS securities score 276 -0.0009 0.0208 -0.1210 0.0857 -0.0104 0.0000 0.0102 contribution / GSIB score Δ level 3 assets score contribution / 270 -0.0011 0.0287 -0.1935 0.1791 -0.0085 -0.0010 0.0081 GSIB score Δ cross-jurisdictional claims score 276 0.0006 0.0113 -0.0489 0.0622 -0.0048 0.0004 0.0054 contribution / GSIB score Δ cross-jurisdictional liabilities 276 0.0013 0.0184 -0.1096 0.0806 -0.0066 0.0015 0.0089 score contribution / GSIB score Δ GSIB score / GSIB score 276 0.0464 0.1132 -0.3099 0.3864 -0.0238 0.0317 0.1062 Δ SRISK / GSIB score 264 10.898 90.896 597.037 505.002 32.344 10.208 51.779