## wp18222 ## Source details **Canonical URL:** [wp18222](https://www.imf.org/-/media/files/publications/wp/2018/wp18222.pdf) ## Other formats - [Markdown version](/-/media/files/publications/wp/2018/wp18222.pdf.md) - [Structured JSON version](/-/media/files/publications/wp/2018/wp18222.pdf.json) --- ### Introduction — context and research question - Motivation: Post-2008 Global Financial Crisis debate on costs and benefits of higher bank capital under Basel III; contribution provides novel bank-level evidence from an emerging market (Peru) where bank-specific, time-varying capital requirements were introduced gradually during 2009–2016. - Key tradeoffs highlighted: - Higher capital requirements can improve loss-absorbing capacity and reduce leverage pro-cyclicality. - Banks meeting requirements by shrinking assets may reduce lending availability and raise lending costs. - Empirical focus: Transitory effects of capital requirement increases on bank lending using Peruvian bank-specific reforms and quarterly bank-level data. ### Data, Peruvian reform design, and identification - Sample and data: - Capital requirement and balance sheet data for all 16 Peruvian commercial banks, period 2005–2016; final sample: 14 banks (after exclusions), accounting for about 85 percent of financial system assets in 2016. - Quarterly frequency; loan growth winsorized at first and 99th percentile. Quarterly loan growth averaged 3.2 percent in real terms (consistent with aggregate 13.7 percent real annual growth). - Reform design and timeline: - Phase 1 (uniform minimum): minimum capital requirement raised from 9.1% to 10% of RWA; announced July 2008; implemented July 1 of 2009, 2010, and 2011; an informal capital adequacy requirement of 11.1 percent was enforced until the beginning of the second reform. - Phase 2 (bank-specific buffers): formula-based, announced July 2011; implemented July 1 of 2012–2016 with required holdings of 40, 55, 70, 85 and 100 percent of the formula-prescribed buffer on July 1 of the years 2012 to 2016, respectively. - Maximum buffer size: buffers could be as high as 5.6 percentage points depending on bank characteristics. - Buffer components: non-cyclical sectoral and regional concentration risk buffers (Herfindahl-Hirschman indices over 19 sectors and 8 regions), top-20 exposures buffer (> 5 percent trigger), risk-propensity buffer (loan provisions and RWAs over last five years; affected only 4 banks), small systemic buffer for the four largest banks, and a countercyclical capital buffer (CCyB) activated until end-October 2014 then deactivated November 2014. - CCyB release rules: countercyclical buffers are only released once a bank has used all its countercyclical provisions; then only 60 percent of the accumulated buffer becomes available; using the remaining 40 percent requires regulatory approval. - Identification strategy: - Exploits bank-specific variation in buffer "jumps" on July 1 of 2012–2016 (annual step-ups) and uses bank and time fixed effects; monthly "wiggles" are not used for identification to mitigate endogeneity. - Controls: log total assets, liquidity = liquid assets / total assets, profitability = ROA, RWA / total assets, excess capital = CAR minus KR. - Implicit assumption: changes in credit demand are common across banks (or controlled). ### Main empirical findings (quarterly bank-level analysis) - Contemporaneous effect: - A one percentage point increase in required capital buffers is associated with a reduction in loan growth of between 4 and 6 percentage points in the same quarter (baseline OLS estimates). - Specification estimates reported roughly -0.0399 to -0.0643 (i.e., -3.99 to -6.43 percentage points) depending on specification and weighting. - Short- to medium-term dynamics: - The contemporaneous negative effect "washes out" quickly: for straddles of six months and beyond, the coefficient on the change in requirements is not statistically different from zero. - Local projection impulse response: effect on cumulative credit growth is limited to a single quarter. - Joint effects of leads and lags are never statistically different from zero: the sum of coefficients on lags and forwards is not significantly different from zero. - Heterogeneity by profitability: - Less profitable banks (below median ROA) contemporaneously reduced quarterly loan growth by -0.0738 (i.e., -7.38 percentage points) unweighted; -0.0649 weighted in some specifications. - More profitable banks (above median ROA) show contemporaneous reductions around -0.0489 (i.e., -4.89 percentage points) unweighted; weighted estimates approx. -0.0642 in some specifications. - Cutting at the 25th percentile produces statistically different coefficients in unweighted regressions but not in weighted regressions. - Loan-type and product-level evidence: - CCyB was implemented with product-specific risk-weight buckets (additional risk-weights 0–55 percent across eight buckets; example: mortgage loans in 15 percent bucket). - Product-level regressions confirm negative impact on lending does not extend beyond one quarter. - Weighted regression of annual credit growth shows a marginally significant positive coefficient on capital surcharge, possibly reflecting product-specific demand changes. ### Robustness and endogeneity checks - Robustness to alternative measures and samples: - Using net rather than gross exposures, Davis and Haltiwanger (1992) measure, restricting sample to 2010–2016, and assuming 11.1 percent minimum until July 2012 produce stable contemporaneous estimates. - Robustness range: contemporaneous drop in lending of between 3 and 6 percentage points per one percentage point increase in capital requirements across specifications. - Endogeneity and anticipation: - Virtual buffers (counterfactual assuming 40 percent implementation before July 1, 2012) show no downward trend between announcement (June 2011) and implementation, suggesting limited strategic balance-sheet adjustments by banks. - Visual inspection and formal tests do not reject equality of averages before and after the announcement; wiggles did not exhibit a downward trend until CCyB switch-off at end-October 2014. - Panel bank-product regressions leverage exogenous surcharge sizes and CCyB activation/deactivation timing to further mitigate endogeneity concerns. - Product-level regression evidence (selected coefficients from Table A5): - ΔCS t, t-1 estimates include -0.0952 (0.068), -0.0389** (0.019), -0.0952 (0.065), -0.0222 (0.017), -0.1257 (0.078), -0.0134 (0.020), -0.0105 (0.092), 0.0008 (0.022) across columns. - Cumulated effect examples: -0.0591 (p-value 0.696), 0.0545 (p-value 0.158), -0.242 (p-value 0.388), 0.0868 (p-value 0.293), 0.418 (p-value 0.267), 0.164 (p-value 0.182). - Significance legend: *** p<0.01, ** p<0.05, * p<0.1. ### Conceptual interpretation, necessary conditions, and caveats - Necessary conditions for capital requirements to affect lending: - (i) requirements must be essentially binding, (ii) additional capital must be costly relative to debt, and (iii) credit demand must be somewhat elastic. - Peruvian data: requirements were essentially binding for most banks (positive correlation between regulatory capital holdings and capital requirements). - Interpretation of temporary effect: - Observed reduction in lending is short-lived (about one quarter); after half a year and beyond, loan growth is statistically indistinguishable from no-change. - Possible explanations for rapid adjustment: low economic cost of raising additional capital for Peruvian banks; high profitability enabling retained earnings to absorb requirement increases; early announcement and slow multi-year implementation allowing smooth adjustment. - Caveats limiting external validity: - Reforms occurred during a growing economy with high bank profitability: between 2011 and 2016, GDP grew at an average annual rate of 4.6 percent; banks’ return on equity averaged 16 percent (non-weighted) and 20 percent (asset-weighted). - Buffer requirements have less severe regulatory consequences than minimum capital requirement breaches; 6 out of 14 banks in sample breached buffer requirements at some point. Therefore estimates based on buffer changes likely constitute a lower bound for effects of equivalent increases in minimum capital requirements. - MM irrelevance and adjustment costs: - If Modigliani and Miller (MM) irrelevance holds, permanent loan supply effects should be absent; adjustment costs can produce transitory credit contractions even if long-run effects are null. - Stylized decomposition: permanent reduction = γ ΔKR; transitory adjustment cost = β ΔKR; observed short-lived contraction consistent with significant β and negligible γ in Peru. ### Key statistics and sample summary (preserved exact values) - Comparative 2012–16 averages: - GDP per capita in Peru: US$5,828 - Average MIC GDP per capita: US$4,744 - Average HIC GDP per capita: US$35,702 - Domestic credit to private sector in Peru: 33 percent of GDP - Average MIC domestic credit to private sector: 46 percent of GDP - Average HIC domestic credit to private sector: 91 percent of GDP - Bank capital over total assets in Peru: 11.5 percent - Bank capital over total assets in MICs: 10.5 percent - Bank capital over total assets in HICs: 8.8 percent - Returns on equity in Peru and MICs: 13 percent - Returns on equity in HICs: 8 percent - Table A3 summary statistics (selected): - ΔL t, t-1: Mean 3.161 | Median 2.671 | Std. Deviation 5.503 | Minimum -10.645 | Maximum 23.315 | Sample size 5,593 - ΔKR t, t-1: Mean 0.717 | Median 0.480 | Std. Deviation 0.507 | Minimum 0.198 | Maximum 2.553 | Sample size 370 - CAR minus KR, t-1: Mean 3.543 | Median 3.044 | Std. Deviation 2.564 | Minimum -0.959 | Maximum 19.109 | Sample size 6,616 - Log Assets, t-1: Mean 15.474 | Median 15.223 | Std. Deviation 1.418 | Minimum 12.236 | Maximum 18.444 | Sample size 6,07 - ROA, t-1: Mean 0.551 | Median 0.486 | Std. Deviation 0.429 | Minimum -1.393 | Maximum 3.184 | Sample size 6,02 - Liquid / Total Assets, t-1: Mean 20.671 | Median 19.572 | Std. Deviation 9.662 | Minimum 3.374 | Maximum 53.012 | Sample size 6,07 - RWA / Total Assets, t-1: Mean 82.880 | Median 84.341 | Std. Deviation 14.143 | Minimum 38.280 | Maximum 160.748 | Sample size 6,07 - Peru: Bank profitability (annual averages across banks, ROA and ROE): - ROA (Unweighted): 2011: 2.16% | 2012: 3.48% | 2013: 1.77% | 2014: 1.75% | 2015: 2.65% | 2016: 1.78% - ROA (Weighted): 2011: 2.26% | 2012: 2.06% | 2013: 1.90% | 2014: 1.88% | 2015: 1.95% | 2016: 2.03% - ROE (Unweighted): 2011: 18.09% | 2012: 19.89% | 2013: 13.98% | 2014: 13.19% | 2015: 15.38% | 2016: 13.35% - ROE (Weighted): 2011: 22.78% | 2012: 21.24% | 2013: 20.04% | 2014: 19.07% | 2015: 20.46% | 2016: 18.48% - System concentration (Table A1): top four banks’ shares (end-2016): - Banco de Crédito del Perú (BCP): Share 34.52% - BBVA (ex Banco Continental): Share 20.69% - Scotiabank Perú: Share 16.08% - Interbank: Share 11.75% - Cumulative share top 4: 83.04% ### Policy-relevant conclusions and recommendations - Empirical conclusion: - A one percentage point increase in required capital buffers reduced quarterly loan growth by between 4 and 6 percentage points in the quarter the increase came into effect, but this effect dissipated quickly and was not detectable over six months or longer. - Policy implications under Peruvian conditions (early announcement, gradual phase-in, high bank profitability): - Higher capital buffers can be implemented with only temporary disruptions to credit supply. - Estimates based on buffer changes should be interpreted as lower bounds for the impact of equivalent increases in minimum capital requirements because buffer breaches have milder consequences. - Heterogeneity matters: less profitable and smaller banks are more sensitive to buffer increases; targeted transitional measures may be warranted to avoid disproportionate short-term credit contractions in weaker institutions. - Design recommendations implied by the Peruvian experience: - Announce reforms well in advance of implementation. - Phase in increases gradually (multi-year step-ups) to allow retained earnings and balance-sheet adjustments to blunt credit disruptions. - Monitor bank profitability and heterogeneity to calibrate complementary measures (e.g., transitional support for less profitable banks) if safeguarding credit provision is a priority. *Source: wp18222 - References* ### References ............................................................................................................. ### References ### Glossary - CAR Capital Adequacy Ratio - CCyB Countercyclical Capital Buffer - FSAP Financial Sector Assessment Program - FSB Financial Stability Board - HIC High-Income Country - MIC Middle-Income Country - MM Modigliani and Miller - RWA Risk-weighted Assests - SBS Superintendencia de Banca y Seguros ### Introduction — context and research question - The 2008 Global Financial Crisis (GFC) prompted higher bank capital requirements under Basel III, raising debate on costs and benefits of bank capital (see Aiyar et al. 2015; Dagher et al. 2016). - Tradeoffs highlighted: - Higher capital requirements can improve loss-absorbing capacity and reduce leverage pro-cyclicality (Admati and Hellwig 2014). - If banks meet requirements by shrinking assets, lending availability and cost may be affected (Kashyap et al., 2010; Hanson et al. 2011). - The Financial Stability Board (FSB) launched the “Framework for Post-Implementation Evaluation of the Effects of the G20 Financial Regulatory Reforms” (FSB, 2017). - Contribution: provides novel evidence from an emerging market—Peru—where bank-specific, time-varying capital requirements were introduced gradually during 2009–2016. ### Data, setting, and motivation for Peru - Peru is relevant because results from advanced economies may not generalize to emerging markets given differences in economic development, credit market competition, and bank performance. - Comparative statistics (2012–16 averages from the Global Financial Development Database): - GDP per capita in Peru: US$5,828 - Average MIC GDP per capita: US$4,744 - Average HIC GDP per capita: US$35,702 - Domestic credit to private sector in Peru: 33 percent of GDP - Average MIC domestic credit to private sector: 46 percent of GDP - Average HIC domestic credit to private sector: 91 percent of GDP - Bank capital over total assets in Peru: 11.5 percent - Bank capital over total assets in MICs: 10.5 percent - Bank capital over total assets in HICs: 8.8 percent - Returns on equity in Peru and MICs: 13 percent - Returns on equity in HICs: 8 percent - The Peruvian reform design (gradual, bank-specific buffers) and rich bank-level data facilitate empirical identification of capital requirement effects on lending. ### Main empirical findings (Peru, quarterly bank-level analysis) - Size of buffers: - Bank-specific additional capital requirements could be as high as 5.6 percentage points depending on bank characteristics. - Short-run impact: - A one percentage point increase in capital requirements is associated with a reduction in loan growth of 4 to 6 percentage points in the same quarter, controlling for bank characteristics and fixed effects. - The negative impact is short-lived and becomes statistically insignificant in about half a year. - Medium/long run: - Over periods longer than a quarter, the effect on cumulative credit growth is statistically insignificant. - Heterogeneity: - Less profitable banks appear more sensitive to changes in capital requirements than more profitable banks. - Contrary to some prior studies, there is no consistent evidence that banks retrench more from loans with higher risk weights in Peru. - Robustness: - Results are robust to different specifications addressing exogeneity concerns and reform anticipation. ### Comparison with existing literature - Effect size and duration: - The short-lived negative effect in Peru is broadly consistent with literature from advanced economies. - Magnitude similar to Aiyar et al. (2014), though the Peru effect is shorter-lived. - Closer to Noss and Toffano (2016), who find effects fade to zero in about a year. - Steady-state vs transitory effects: - Literature distinguishes transitory (adjustment period) and steady-state (after full convergence) effects. - Estimating steady-state effects is challenging due to confounding factors and endogeneity; many studies rely on model calibration and the Modigliani and Miller (MM) offset. - Calibration literature on MM offset: - Most studies allowing for MM offset find small effects of capital on lending rates; a 1 percentage point higher requirement often implies about a 2 basis points increase in rates (Kashyap et al., 2010; Miles et al., 2013). - Studies assuming limited MM offset estimate up to 13 basis points increase in rates (Elliott, 2009; BCBS, 2010). - Transitory-effect literature: - Peek and Rosengren (1995): banks under capital conservation plans lent 2 percent less over one year. - Aiyar et al. (2014): a 1 percentage point increase in requirement associated with 5.7–8 percent lower lending over the following three quarters. - Noss and Toffano (2016): a 15 basis points increase in requirements leads to a 0.25 percentage point reduction in quarterly lending growth after two quarters, fading to zero after about one year. - Fraisse et al. (2017): a 1 percentage point higher requirement associated with 9 percent lower lending in the following year. - Mesonnier and Monks (2014): a 1 percentage point higher capital shortfall associated with 1.6 percent lower lending in the following year. - Heterogeneous loan-type effects: - Studies show stronger retrenchment for riskier or arm’s-length lending (e.g., Imbierowicz et al. 2018; Bridges et al. 2014; Gropp et al. 2018; De Jonghe et al. 2016; Behn et al. 2016). ### Peruvian capital reforms (timeline and key design features) - Two-phase reform by SBS: - 2009–2011: uniform minimum requirements raised from 9.1 percent to 10 percent of RWA. - 2012–2016: introduction of bank-specific capital buffers on top of the 10 percent uniform minimum; buffers could be as high as 5.6 percentage points. - The first reform, announced July 2008, stipulated that on July 1 of 2009, 2010, and 2011, uniform minimum capital requirements would be raised from 9.1 percent to 9.5 percent (as part of the staged increase toward 10 percent). ### Conceptual and methodological approach - The paper analyzes transitory effects using a bank-level lending regression with bank and time fixed effects to absorb bank-specific unobservables and common shocks. - A simple conceptual framework is introduced to analyze the effect of capital requirements on credit, distinguishing between banks satisfying requirements by retaining earnings versus shrinking assets. - The empirical strategy exploits the timing and bank-specific nature of Peru’s capital buffers for identification. ### Policy-relevant implications (as reflected in findings) - Gradual, anticipated implementation and high bank profitability can reduce the contractionary impact of higher capital requirements by enabling banks to meet requirements via retained earnings. - Short-lived negative effects suggest that temporary credit adjustments during transition to higher capital may be limited in expanding economies with profitable banks. - Heterogeneous sensitivity by profitability indicates targeted supervisory attention may be warranted for less profitable banks during capital requirement increases. *Source: wp18222 - References* ### 9.8 percent to of RWA, respectively. Instead of introducing specific numerical increases, the ### wp18222 - 9.8 percent to of RWA, respectively. Instead of introducing specific numerical increases, the ### Reform design and implementation - Phase 1 (raising uniform requirements): - Minimum capital requirement raised from 9.1% to 10% of RWA. - Announced in ’08; implemented July ‘09–’11. - An informal capital adequacy requirement of 11.1 percent was enforced until the beginning of the second reform. - Phase 2 (bank-specific buffers): - Formula-based, announced in July ’11; implemented July ‘12–’16. - Implementation schedule: banks had to hold 40, 55, 70, 85 and 100 percent of the formula-prescribed buffer on July 1 of the years 2012 to 2016, respectively. - Buffer amounts were recalculated monthly (producing “wiggles”), while annual step-ups on July 1 produced “jumps”. - Compulsory capital buffers consisted of: - Non-cyclical component: sectoral and regional concentration risk buffers (Herfindahl-Hirschman indices over 19 sectors and 8 regions), top-20 exposures buffer (kicks in when > 5 percent), risk-propensity buffer (driven by loan provisions and RWAs over last five years; affected only 4 banks in sample), and a small systemic buffer for the four largest banks. - Countercyclical capital buffer (CCyB): born alive and remained switched on until end of October 2014; deactivated November ’14. When CCyB was switched off in November, no additional CCyB had to be accumulated against new lending, but already accumulated buffers did not become available and had to be topped up to 85 percent and 100 percent of the formula on July 1 of 2015 and 2016, respectively. - In Peru, countercyclical buffers are only released once a bank has used all its countercyclical provisions. Even then, only 60 percent of the accumulated buffer becomes available; using the remaining 40 percent requires regulatory approval. ### Data, sample, and identification strategy - Data: - Capital requirement and balance sheet data for all 16 Peruvian commercial banks, period 2005–2016; source: SBS. - Transformed to quarterly frequency; winsorized loan growth at the first and 99th percentile. - Sample construction: - Two banks (Cencosud and ICBC) dropped due to short history; newly-established banks excluded observations for first 4 quarters. - Final sample: 14 banks, accounting for about 85 percent of financial system assets in 2016. - Quarterly loan growth averaged 3.2 percent in real terms (consistent with aggregate 13.7 percent real annual growth). - Identification: - Focus on the second reform due to bank-specific variation in required buffers enabling difference-in-differences with bank and time fixed effects. - Use only annual “jumps” (July 1 step-ups) for identification; do not use monthly “wiggles” to mitigate endogeneity. - Controls: (1) log total assets, (2) liquidity = liquid assets / total assets, (3) profitability = ROA, (4) RWA / total assets, (5) excess capital = CAR minus KR. - Implicit assumption: changes in demand for credit are the same for all banks (and for bank-product regressions, that product-specific demand shocks are common or controlled). ### Main empirical findings - Contemporaneous effect: - A one percentage point increase in required capital buffers is associated with a reduction in loan growth of between 4 and 6 percentage points in the same quarter (baseline OLS estimates). - Table 1 and Table 4 (specification variations) consistently show a negative contemporaneous effect of roughly -0.0399 to -0.0643 (i.e., -3.99 to -6.43 percentage points) depending on specification and weighting. - Short- to medium-term dynamics: - The contemporaneous negative effect “washes out” quickly: for straddles of six months and beyond, the coefficient on the change in requirements is not statistically different from zero. - Local projection impulse response confirms effect on cumulative credit growth is limited to a single quarter. - Joint effects of leads and lags are never statistically different from zero: the sum of coefficients on lags and forwards is not significantly different from zero. - Heterogeneity by profitability: - Less profitable banks (below median ROA) show a larger contemporaneous reduction in quarterly loan growth: -0.0738 (i.e., -7.38 percentage points) unweighted; -0.0649 weighted in some specifications. - More profitable banks (above median ROA) show a contemporaneous reduction around -0.0489 (i.e., -4.89 percentage points) unweighted; weighted estimates approx. -0.0642 in some specifications. - When cutting at the 25th percentile, coefficients differ statistically in unweighted regressions but not in weighted ones. - Product-level (CCyB) evidence addressing endogeneity: - CCyB is product-specific via eight buckets (additional risk-weights 0–55 percent; example: mortgage loans in 15 percent bucket). - Product-level regressions confirm negative impact on lending does not extend beyond one quarter. - Weighted regression of annual credit growth shows marginally significant positive coefficient on capital surcharge, possibly reflecting product-specific demand changes. - Endogeneity checks: - Virtual buffers (counterfactual buffers assuming 40 percent implementation from before July 1, 2012) show no downward trend between announcement (June 2011) and implementation, suggesting limited strategic balance-sheet adjustments by banks to reduce buffer requirements. - Visual inspection and formal tests do not reject equality of averages before and after the announcement; wiggles did not exhibit a downward trend until CCyB switch-off at end-October 2014. - Panel bank-product regressions leverage exogenous surcharge sizes and timing of CCyB activation/deactivation to further mitigate endogeneity concerns. ### Robustness - Alternative loan growth measures: - Using net rather than gross exposures, Davis and Haltiwanger (1992) measure, restricting sample to 2010–2016, and assuming 11.1 percent minimum until July 2012 produce stable contemporaneous estimates. - Robustness results indicate contemporaneous drop in lending of between 3 and 6 percentage points per one percentage point increase in capital requirements across specifications. - Additional specification checks: - Allowing for up to one year of leads and lags in quarterly regressions does not reveal a statistically significant persistent effect; contemporaneous negative effect remains but overall cumulative effect is insignificant. ### Conceptual interpretation and caveats - Necessary conditions for capital requirements to affect lending: - (i) requirements must be essentially binding, (ii) additional capital must be costly relative to debt, and (iii) credit demand must be somewhat elastic. - Peruvian data suggest requirements were essentially binding for most banks (positive correlation between regulatory capital holdings and capital requirements). - Interpretation of temporary effect: - The reduction in lending appears short-lived (about one quarter); after half a year and beyond, loan growth statistically indistinguishable from no-change scenario. - Possible explanations for quick adjustment: - Low economic cost of raising additional capital for Peruvian banks. - High profitability enabling retained earnings to absorb requirement increases. - Early announcement and slow multi-year implementation allowing smooth adjustment. - Two caveats limiting external validity: - Reforms occurred during a growing economy with high bank profitability: between 2011 and 2016, GDP grew at an average annual rate of 4.6 percent, banks’ return on equity averaged 16 percent (non-weighted) and 20 percent (asset-weighted). - Buffer requirements have less severe regulatory consequences than minimum capital requirement breaches; 6 out of 14 banks in sample breached buffer requirements at some point. Hence estimates based on buffer changes likely constitute a lower bound for effects of equivalent increases in minimum capital requirements. - MM irrelevance and adjustment costs: - If MM holds (capital structure irrelevant), permanent loan supply effects should be absent; however, adjustment costs can produce transitory credit contractions even if long-run effects are null. - Stylized decomposition: permanent reduction = γ ΔKR; transitory adjustment cost = β ΔKR; observed short-lived contraction consistent with significant β and negligible γ in Peru. ### Policy-relevant conclusions - A one percentage point increase in required capital buffers reduced quarterly loan growth by between 4 and 6 percentage points in the quarter the increase came into effect, but this effect dissipated quickly and was not detectable over six months or longer. - The short-lived impact suggests that, under conditions of early announcement, gradual implementation, and high bank profitability: - Higher capital buffers can be implemented with only temporary disruptions to credit supply. - Estimates based on buffer changes should be interpreted as lower bounds for the impact of equivalent increases in minimum capital requirements, due to weaker enforcement and milder consequences of buffer breaches. - Heterogeneity matters: less profitable and smaller banks are more sensitive to buffer increases, implying targeted transitional measures may be warranted to avoid disproportionate short-term credit contractions in weaker institutions. - Design recommendations implied by the Peruvian experience: - Announce reforms well in advance of implementation. - Phase in increases gradually (multi-year step-ups) to allow retained earnings and balance-sheet adjustments to blunt credit disruptions. - Monitor bank profitability and heterogeneity to calibrate complementary measures (e.g., transitional support for less profitable banks) if safeguarding credit provision is a priority. *Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18222.pdf* ### REFERENCES ### wp18222 - REFERENCES ### Major referenced themes - Literature on bank capital regulation, costs and benefits, and macroprudential policy: - Bank capital theory and empirical evidence (Admati and Hellwig (2013); Aiyar, Calomiris, and Wieladek (2015); Miles, Yang, and Marcheggiano (2013); Modigliani and Miller (1958)). - Macroprudential regulation and capital requirements’ effects on lending, risk-taking, and monetary policy transmission (Aiyar et al. (2014); Behn et al. (2016); Dagher et al. (2016); Gambacorta and Shin (2016); Hanson, Kashyap, and Stein (2011)). - Empirical studies on capital requirement implementation and bank responses (Bridges et al. (2014); Gropp et al. (2018); Noss and Toffano (2016); Fraisse, Le, and Thesmar (2017)). - Measurement and methodological contributions (Jorda (2005); Kisin and Manela (2016); Karan and Zinman (2013)). ### Appendix I — Data description (Peru) - Table A1: Commercial Banks in Peru as of end-2016 (selected entries, total assets in thousands of Soles, rank, share, cumulative share) - Banco de Crédito del Perú (BCP): 121,913,255 | Rank 1 | Share 34.52% | Cumulative share 34.52% - BBVA (ex Banco Continental): 73,068,119 | Rank 2 | Share 20.69% | Cumulative share 55.21% - Scotiabank Perú: 56,792,357 | Rank 3 | Share 16.08% | Cumulative share 71.29% - Interbank: 41,482,405 | Rank 4 | Share 11.75% | Cumulative share 83.04% - Banco Interamericano de Finanzas (BANBIF): 12,858,287 | Rank 5 | Share 3.64% | Cumulative share 86.68% - Mibanco: 11,509,224 | Rank 6 | Share 3.26% | Cumulative share 89.93% - Banco Financiero: 8,780,702 | Rank 7 | Share 2.49% | Cumulative share 92.42% - Citibank: 5,975,162 | Rank 8 | Share 1.69% | Cumulative share 94.11% - Banco GNB: 5,289,726 | Rank 9 | Share 1.50% | Cumulative share 95.61% - Banco Falabella Perú: 4,867,627 | Rank 10 | Share 1.38% | Cumulative share 96.99% - Banco Santander Perú: 4,783,806 | Rank 11 | Share 1.35% | Cumulative share 98.34% - Banco Ripley: 2,238,754 | Rank 12 | Share 0.63% | Cumulative share 98.98% - Banco de Comercio: 1,945,597 | Rank 13 | Share 0.55% | Cumulative share 99.53% - Banco Cencosud: 681,960 | Rank 14 | Share 0.19% | Cumulative share 99.72% - Banco Azteca del Perú: 563,288 | Rank 15 | Share 0.16% | Cumulative share 99.88% - Banco ICBC: 421,146 | Rank 16 | Share 0.12% | Cumulative share 100.00% ### Variable definitions and units (Table A2) - ΔL t, t-1: Quarter-on-quarter change in real gross loans. Constructed from end-of-quarter loan levels deflated with consumer prices. Unit: percent - ΔKR t, t-1: Quarter-on-quarter "jumps" in capital buffer requirement, expressed as percent of RWA. Jumps equal zero except at step-ups on July 1 of the years 2012 to 2016. Constructed from averages of monthly-levels data in each quarter. Unit: percent - CAR minus KR, t-1: Difference between total regulatory capital held and total capital requirement during quarter t-1, both expressed as percent of RWAs. Unit: percent - Log Assets, t-1: Natural logarithm of total assets on balance sheet at the end of quarter t-1. Unit: log of Soles - ROA, t-1: Return On Assets during quarter t-1, defined as Net Income divided by Total Assets. Unit: percent - Liquid / Total Assets, t-1: Ratio of liquid assets and total assets at the end of quarter t-1. Unit: percent - RWA / Assets, t-1: Ratio of risk-weighted assets and total assets at the end of quarter t-1. Unit: percent ### Summary statistics (Table A3, Table A4) - Summary statistics for key variables (all changes and ratios in percent). Sample sizes reported. - ΔL t, t-1: Mean 3.161 | Median 2.671 | Std. Deviation 5.503 | Minimum -10.645 | Maximum 23.315 | Sample size 5,593 - ΔKR t, t-1: Mean 0.717 | Median 0.480 | Std. Deviation 0.507 | Minimum 0.198 | Maximum 2.553 | Sample size 370 - CAR minus KR, t-1: Mean 3.543 | Median 3.044 | Std. Deviation 2.564 | Minimum -0.959 | Maximum 19.109 | Sample size 6,616 - Log Assets, t-1: Mean 15.474 | Median 15.223 | Std. Deviation 1.418 | Minimum 12.236 | Maximum 18.444 | Sample size 6,07 - ROA, t-1: Mean 0.551 | Median 0.486 | Std. Deviation 0.429 | Minimum -1.393 | Maximum 3.184 | Sample size 6,02 - Liquid / Total Assets, t-1: Mean 20.671 | Median 19.572 | Std. Deviation 9.662 | Minimum 3.374 | Maximum 53.012 | Sample size 6,07 - RWA / Total Assets, t-1: Mean 82.880 | Median 84.341 | Std. Deviation 14.143 | Minimum 38.280 | Maximum 160.748 | Sample size 6,07 - Note: All changes and ratios expressed in percent. ΔKR t, t-1 only accounts for non-zero values (implementation periods only). ROA, t-1 is on a quarterly basis. - Peru: Bank Profitability (Table A4) — Profitability during 2011–2016. Annual averages across banks: - ROA (Unweighted): 2011: 2.16% | 2012: 3.48% | 2013: 1.77% | 2014: 1.75% | 2015: 2.65% | 2016: 1.78% - ROA (Weighted): 2011: 2.26% | 2012: 2.06% | 2013: 1.90% | 2014: 1.88% | 2015: 1.95% | 2016: 2.03% - ROE (Unweighted): 2011: 18.09% | 2012: 19.89% | 2013: 13.98% | 2014: 13.19% | 2015: 15.38% | 2016: 13.35% - ROE (Weighted): 2011: 22.78% | 2012: 21.24% | 2013: 20.04% | 2014: 19.07% | 2015: 20.46% | 2016: 18.48% ### Product-level regressions and key estimates (Table A5) - Dependent variable: ΔL t, t-1 (quarterly loan growth). Estimates show the effect of a 1 p.p. jump in the capital surcharge (ΔCS) allowing for lags and leads up to one year. - Selected coefficient estimates and standard errors (robust standard errors in parentheses): - ΔCS t, t-1: Columns show varied estimates: - (1) -0.0952 (0.068) - (2) -0.0389** (0.019) - (3) -0.0952 (0.065) - (4) -0.0222 (0.017) - (5) -0.1257 (0.078) - (6) -0.0134 (0.020) - (7) -0.0105 (0.092) - (8) 0.0008 (0.022) - ΔCS t-1, t-2: - (2) 0.0373** (0.018) - (4) 0.0460** (0.023) - (6) 0.0349* (0.021) - ΔCS t-2, t-3: - (2) -0.0321* (0.019) - ΔCS t+1, t: - (2) 0.0393* (0.020) - (4) 0.0480** (0.022) - (6) 0.0586** (0.027) - ΔCS t+3, t+2: - (1) 0.1641* (0.098) - (2) 0.0388** (0.018) - Other control variable coefficients (examples, standard errors in parentheses): - CAR minus KR, t-1: coefficient estimates range from -0.6108 to 0.5116 across columns with standard errors ranging 0.330 to 1.185. - Log Assets, t-1: coefficients around -0.0424 to -0.0250 with standard errors 0.041 to 0.121. - ROA, t-1: coefficients show large variation (e.g., -16.6507 (14.222), 2.6005 (3.904)). - Regression details: - Observations by column: (1) 2,741 | (2) 2,741 | (3) 2,626 | (4) 2,626 | (5) 2,511 | (6) 2,511 | (7) 2,292 | (8) 2,292 - R-squared by column: 0.043 | 0.159 | 0.056 | 0.159 | 0.054 | 0.167 | 0.071 | 0.176 - Product x Bank fixed effects: Yes (all columns) - Time fixed effects: Yes (all columns) - Standard errors: Cluster (all columns) - Weights: Columns alternate No/Yes - Cumulated effect estimates and p-values reported: - Cumulated effect: -0.0591 (p-value 0.696), 0.0545 (p-value 0.158), -0.242 (p-value 0.388), 0.0868 (p-value 0.293), 0.418 (p-value 0.267), 0.164 (p-value 0.182) - Significance legend: *** p<0.01, ** p<0.05, * p<0.1 ### Figures (descriptions) - Figure A1: Peru: Actual Versus Required Capital — Actual regulatory capital shown in grey, required capital shown in blue. - Figure A2: Peru: Impulse Response Function for Product-level Capital Surcharges — Response of cumulative credit growth to a 1 pp. rise in capital surcharge, assuming rise and “shock” both occur at time zero. Estimates based on unweighted regressions with all controls of Table 5 in the main text. *Source: wp18222 - REFERENCES (Appendix I content and references list).* --- _Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18222.pdf_