## wp17245

## Source details

**Canonical URL:** [wp17245](https://www.imf.org/-/media/files/publications/wp/2017/wp17245.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp17245.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp17245.pdf.json)

---

### I. Introduction — objectives and contributions
- Objective: analyze the impact of capital and liquidity on bank-lending following the 2008 financial crisis, incorporating Basel III–inspired measures of capital and liquidity.
- Key research questions:
  - Whether the role of bank capital on lending differs when using risk-weighted versus leverage regulatory capital ratios.
  - Whether the role of bank liquidity on lending differs when using liquidity measures inspired by Basel III.
- Sample and approach:
  - Unbalanced panel of commercial banks over 2008–15 with focus on U.S. and European banks.
  - Balance sheets converted to IFRS where needed; estimation period set at 2009 to capture anticipation and adaptation to Basel III.
- Main contribution: assesses capital and liquidity impacts on bank-lending in the post-2008 period using new Basel III–inspired measures.

### II. Data, sample, and measurement
- Coverage and sources:
  - Commercial banks from 23 countries over 2008–15; data from S&P Global Market Intelligence and Bloomberg; macro indicators from DataStream and IMF World Economic Outlook Database.
- Sample construction:
  - Initially 1,341 commercial banks (1,040 in the United States and 301 in Europe) for 2008–15.
  - Liquidity indicators calculated only for banks with loan-category and deposit-maturity breakdowns.
  - Banks with regulatory capital ratios below regulatory minimums were deleted.
  - Final sample: 1,058 commercial banks (789 in the United States and 269 in Europe).
- Representativeness:
  - Final sample accounts, on average, for 81 percent of total loans from U.S. commercial banks and, on average, for 58 percent of total loans from European commercial banks (country variation e.g., 6 percent in the Slovak Republic and 84 percent in Belgium).
- Regulatory minimums (as stated):
  - Minimum requirement at 8 percent for total regulatory capital to total risk-weighted assets (except Cyprus at 10 percent; United Kingdom considered equal to 9 percent following Jokipii and Milne (2008)).
  - Tier 1 capital ratio minimums: 4 percent before 2013, 4.5 percent in 2013, 5.5 percent in 2014, and 6 percent in 2015 (Basel III phase-in arrangements).
  - Ratios of Tier 1 and Core Tier 1 capital to total assets minimum at 3 percent following the “Basel III leverage ratio framework and disclosure requirements” (BIS, 2014a).
- Measurement choices:
  - Aggregate total loans used as dependent variable to improve comparability across accounting standards.

### III. Stylized facts (2008–15)
- Lending growth patterns:
  - U.S. banks: drop in lending-growth over 2008–11; recovery from 2011 stronger for commercial loans than retail-and-other loans.
    - Commercial loans growth rates: 12.8, 3.6, and 0.4 percent for 2008, 2009, 2010 respectively; recovery up to 11.8 percent in 2015.
    - Retail-and-other-loans growth rates: 14.9, 10, 3.4, and 1.6 percent for 2008–11 respectively; recovery up to 8.9 percent in 2015.
  - European banks: continuous downward trends in lending-growth over 2008–15; credit crunch since 2014.
    - Commercial loans growth rate negative since 2014, e.g., -9.6 percent (specific years indicated in source).
    - Retail-and-other-loans growth rates turning negative since 2014, e.g., -9.6 and -6.1 percent for respective years.
- Capital ratios:
  - European banks increased capital ratios more significantly than U.S. banks over 2008–15; U.S. banks have higher leverage ratios.
    - European banks:
      - Total risk-weighted capital ratio: increased from 13.5 to 17.9 percent (2008–15).
      - Tier 1 capital ratio: increased from 10.7 to 15.8 percent.
      - Tier 1 leverage ratio: increased from 6.2 to 7.3 percent.
      - Core Tier 1 leverage ratio: increased from 6.0 to 7.0 percent.
    - U.S. banks:
      - Tier 1 leverage ratio: increased from 10.2 percent in 2008 to 10.4 percent in 2015.
      - Core Tier 1 leverage ratio: increased from 9.6 percent in 2008 to 10.1 percent in 2015.
      - Total risk-weighted capital ratio: increased from 15.5 percent in 2008 to 16.1 percent in 2015, peaking at 17.0 percent in 2012.
      - Tier 1 capital ratio: increased from 14.3 percent in 2008 to 14.8 percent in 2015, peaking at 15.6 percent in 2012.
- Liquidity and stable funding:
  - U.S. banks hold more stable funding than European banks; European banks hold more liquid assets than U.S. banks.
  - Trends:
    - U.S. stocks of liquid assets: 29.9 percent in 2008, 35.5 percent in 2012, 32.0 percent in 2015.
    - Available amount of stable funding (Basel III–inspired) for U.S. banks: 87.8 percent in 2008 to 84.4 percent in 2015.
    - Core deposits for U.S. banks: 75.1 percent in 2008 to 81.5 percent in 2013, then 78.9 percent in 2015.
- Definitions used (selected):
  - Total commercial loans: loans to commercial and industrial entities + commercial real estate loans + construction loans + loans to agriculture + loans to money market funds.
  - Total retail-and-other-loans: credit card loans + installment loans + residential mortgage loans + loans to municipalities and governments + other loans.
  - Total risk-weighted capital ratio: (Tier 1 capital + Tier 2 capital)/total risk-weighted assets.
  - T1 capital ratio: Tier 1 capital/total risk-weighted assets.
  - T1 leverage ratio: Tier 1 capital/total assets.
  - Core Tier 1 leverage ratio: Core Tier 1 capital/total assets.
  - Nonrequired amount of stable funding: nonrequired amount of stable funding (based on Basel III definition)/total assets.
  - Available amount of stable funding: available amount of stable funding (based on Basel III definition)/total assets.
  - Available amount of stable funding (including Core deposits) for U.S. banks: available amount of stable funding (Core deposits being considered as stable)/total assets.
  - Core deposit definition: sum of deposits lower than US$100,000, regardless of maturity.

### IV. Empirical strategy and variables
- Model:
  - 훥LO_i,t = α_i + Σ_j β_j X_ji,t−1 + Σ_k β_k X_ki,t + ε_i,t
  - Dependent variable: lending-growth measured as year-on-year change in loans expressed in current U.S. dollars.
  - Bank-specific variables lagged once (t-1); OLS panel estimator with bank cross-section fixed effects used.
  - Time fixed effects excluded when controlling for demand via macro variables; standard errors robust to heteroskedasticity.
  - All variables winsorized at the 1st and 99th percentile levels.
  - Wooldridge test: null of “no first-order autocorrelation” cannot be rejected; static panel model used.
- Key variables and proxies (preserved exactly):
  - Capital measures: t12k_rwa; t1k_rwa; t1k_ta; ct1k_ta.
  - Liquidity measures: nrasf_ta; aasf_ta; aasf_cd_ta.
  - Credit risk: npl_tlo.
  - Bank risk appetite: roa_sdroa.
  - Funding cost: cost_fund.
  - Profitability: roe.
  - Bank size: ln_ta.
  - Merger/acquisition dummy: DUM_MA (1 if annual growth rate of total assets > 35 percent, 0 otherwise).
  - Macroeconomic variables: ngdp_gwt; 3m_itbnk_c; cb_gdp_gwt.
  - Basel III implementation index: basel3 (varies between 0 and 3).

### V. Main empirical findings — capital effects (U.S. vs Europe)
- U.S. banks:
  - Risk-weighted capital ratios and leverage ratios have significant and positive impacts on bank-lending-growth for both commercial and retail-and-other specifications.
  - Interpretation: capitalization encourages U.S. bank-lending-growth post-2008 as banks strengthen soundness and loss-absorption when expanding credit.
- European banks:
  - Capital ratios not significant determinants of commercial-lending-growth.
  - Tier 1 and Tier 2 risk-weighted capital ratio and Tier 1 leverage ratios have significant and negative impacts on retail-and-other-lending-growth.
  - Interpretation: post-2008 deleveraging and credit crunch; substitution into risk-free government securities in a very low interest-rate environment; possible zombie lending distortions.
- European sample averages:
  - Average risk-weighted capital ratio: 15.9 percent.
  - Average Tier 1 leverage ratio: 6.8 percent.

### VI. Main empirical findings — liquidity effects and perverse outcomes
- nrasf_ta (nonrequired amount of stable funding/total assets):
  - Significant and positive impact on U.S. and European commercial-lending-growth.
  - Significant and negative impact on European retail-and-other-lending-growth.
  - Interpretation: buffer stocks of liquid assets enable expansion in risky, illiquid commercial lending (U.S.), but enable reduction of semiliquid retail loans (Europe) amid asset-shrinking pressures.
- aasf_ta (available amount of stable funding/total assets):
  - Not significant for U.S. and European bank-lending-growth in general.
- aasf_cd_ta (available amount of stable funding with Core deposits for U.S. banks):
  - Significant and negative impact on U.S. retail-and-other-lending-growth for large U.S. banks.
  - Significant and positive impact on U.S. commercial-lending-growth for small U.S. banks.
  - Interpretation: more stable funding may weaken large U.S. banks’ willingness to expand semiliquid retail lending; small U.S. banks rely on Core deposits to expand risky, illiquid commercial lending.

### VII. Heterogeneity by bank size and market access
- Bank-size threshold used in analyses: US$1 billion (large if above US$1 billion; small if below US$1 billion).
- Small U.S. banks:
  - Capital ratios (both risk-weighted and leverage) have significant and positive impacts on lending-growth for both commercial and retail-and-other specifications.
  - Interpretation: capitalization matters more for small banks due to higher costs of raising equity and less reputational advantage.
- Large U.S. banks:
  - Only leverage ratios have significant and positive impacts on commercial-lending-growth.
  - Large banks benefit from broader access to external funding and may allocate less stable funding when expanding semiliquid retail-and-other lending.
- European banks:
  - Large European banks: capital ratios negatively affect retail-and-other-lending-growth; nrasf_ta positive for commercial lending but negative for retail lending.
  - Small European banks: limited ability to reduce commercial loans when holding liquid buffers; coefficients often imprecisely estimated given small N (e.g., N 88).

### VIII. Descriptive statistics and loan shares (selected exact figures)
- Average share of commercial loans in total assets (whole sample): 41.4 percent.
  - Large U.S. banks: 45.4 percent.
  - Small U.S. banks: 43.6 percent.
  - Large European banks: 31.6 percent.
  - Small European banks: 31.9 percent.
- Average share of retail-and-other-loans in total assets (whole sample): 25.3 percent.
  - Large U.S. banks: 21.2 percent.
  - Small U.S. banks: 24.7 percent.
  - Large European banks: 31.2 percent.
  - Small European banks: 37.5 percent.
- Average share of deposits in total assets (whole sample): 74.4 percent.
  - Large U.S. banks: 77.4 percent.
  - Small U.S. banks: 81.5 percent.
  - Large European banks: 53.0 percent.
  - Small European banks: 70.2 percent.
- Bank performance (selected):
  - Small U.S. banks ROE: 3.0 percent.
  - Small U.S. banks nonperforming loans: 3.1 percent.
  - Large U.S. banks ROE: 5.9 percent.
  - Large U.S. banks nonperforming loans: 2.6 percent.

### IX. Robustness checks and methodological notes
- Endogeneity: robustness via GMM using one-year lagged instruments; one-year lags not weak instruments; more lags weak and excluded.
- Robustness outcomes (selected):
  - Some capital effects become not significant under GMM for certain large/small U.S. bank specifications.
  - Negative impact of capital ratios for large European banks weakened under some robustness checks.
  - aasf_cd_ta for U.S. banks becomes significant with negative impact on large U.S. bank-commercial-lending-growth in some GMM specifications.
- Liquidity-weight sensitivity:
  - Consumer loan weight alternatives: 0.15 (nradf015_ta) and 0.5 (nradf05_ta) versus baseline 0.25.
  - Demand and saving deposit weight alternatives: 0.5 (aasf05_ta) and 0.9 (aasf09_ta) versus baseline 0.7.
  - Main conclusions consistent across these alternative weight specifications.
- Subsample checks:
  - Very large banks: total assets above US$50 billion (Federal Reserve “significant” threshold per FED, 2011).
  - Other large banks: total assets between US$50 billion and US$1 billion.
  - Alternative European large threshold: US$20 billion (median); main conclusions consistent.

### X. Selected regression diagnostics and table highlights (examples preserved exactly)
- Table-level diagnostics (examples):
  - U.S. banks commercial-loans regressions: r2 values around 0.221 to 0.225; N values e.g., 5130, 5669.
  - European banks regressions: r2 around 0.194–0.213; N 1595 (various specifications 1595, 1536, 1474).
  - Large U.S. banks: r2 for commercial loans 0.262–0.321; N 1810, 1974, 1949.
  - Small U.S. banks: r2 for commercial loans 0.262–0.308; N 3320, 3695, 3669.
- Appendix coefficient examples (GMM results, selected):
  - Large U.S. banks nrasf_ta: 0.273 (0.214); 0.275 (0.210); 0.369* (0.192) ... 0.596*** (0.174).
  - Small U.S. banks t1k_ta: 1.508*** (0.424); ct1k_ta: 1.446*** (0.428); nrasf_ta: 0.678*** (0.157) ... 0.841*** (0.149).
  - Large European banks t12k_rwa: -0.971** (0.482); nrasf_ta: 0.953*** (0.304) ... 1.348*** (0.350); retail nrasf_ta: -0.333* (0.298) ... -0.661* (0.355).
- Reported statistical significance conventions: * 10 percent; ** 5 percent; *** 1 percent. Student-t statistics or standard errors are reported in parentheses.

### XI. Policy implications and recommendations
- Regulatory trade-offs:
  - Basel III–inspired capital and liquidity standards affect loan composition and funding choices across bank sizes and jurisdictions; trade-offs exist between strengthening banks’ stability and preserving their role as liquidity providers to the real economy.
- Implementation cautions:
  - Uniform implementation of capital and liquidity regulatory frameworks across all banks could have counterproductive effects on post-2008 bank-lending-growth.
  - The effect of new regulatory frameworks may differ depending on economic context and a bank’s ability to originate credit; they may be ineffective or detrimental in some contexts.
- Key recommendation:
  - Regulators need clearer definitions and measurements under a global regulatory framework — specifically, determination of what types of liquid liabilities should be considered stable for a deeper regulatory definition of stable funding.
  - Regulators should consider heterogeneous banks’ characteristics and behaviors when designing and enforcing regulations to promote financial stability while preserving banks’ credit provision functions.

*Source: IMF working paper wp17245 (sections I–IV, hypotheses, empirical strategy, main findings, robustness checks, and Appendix listings as provided in the source content).*

### References .............................................................................................................

### References

### I. Introduction — objectives and contributions
- Objective: analyze the impact of capital and liquidity on bank-lending following the 2008 financial crisis, incorporating new measures of capital and liquidity inspired by the Basel III regulatory framework.
- Key research questions:
  - Whether the role of bank capital on lending differs when using risk-weighted versus leverage regulatory capital ratios.
  - Whether the role of bank liquidity on lending differs when using liquidity measures inspired by Basel III.
- Sample and approach:
  - Empirical evidence for an unbalanced panel of U.S. and European commercial banks over the period 2008–15.
  - New measures of capital and liquidity inspired by Basel III included beyond determinants in existing literature.
  - Balance sheets converted to IFRS accounting standards where needed for consistency.
  - Estimation period set at 2009 to capture banks’ anticipation and adaptation to Basel III implementation.
- Main contributions:
  - Assesses capital and liquidity impacts on bank-lending in the post-2008 period.
  - Applies new Basel III–inspired measures of capital and liquidity.

### II. Main findings and implications
- Aggregate and bank-type differentiated results:
  - Small U.S. banks strengthen financial soundness and loss-absorption capacities when expanding both commercial and retail-and-other-credit activities.
  - Large U.S. banks strengthen leverage ratios when granting riskier, illiquid commercial loans.
  - Capital ratios have significant and negative impacts on bank-retail-and-other-lending-growth for large European banks during the post-2008 deleveraging and “credit crunch.”
  - Liquidity ratios have positive yet perverse effects on bank-lending-growth: U.S. banks hold buffer stocks of liquid assets when expanding risky, illiquid commercial-lending activities.
  - Large European banks have limited ability to reduce commercial loans during credit rationing (2008–15) but can curtail retail-and-other lending while holding buffer stocks of liquid assets.
  - Large U.S. banks access broader external funding and allocate less stable funding when expanding semiliquid retail-and-other-lending activities.
  - Small U.S. banks expand risky and illiquid commercial-lending activities while relying more on stable core deposit funding.

- Policy implication (implicit in findings):
  - Regulatory capital and liquidity standards (Basel III–inspired) affect loan composition and funding choices across bank sizes and jurisdictions; trade-offs exist between strengthening banks’ stability and preserving their role as liquidity providers to the real economy.

### III. Data, sample, and measurement
- Coverage:
  - Commercial banks from 23 countries over 2008–15.
  - Focus on U.S. and European banks; data from S&P Global Market Intelligence and Bloomberg; macro indicators from DataStream and IMF World Economic Outlook Database.
- Sample construction:
  - From 2008 to 2015, 1,341 commercial banks identified (1,040 in the United States and 301 in Europe).
  - Liquidity indicators calculated only for banks with loan-category and deposit-maturity breakdowns available in S&P Global Market Intelligence or Bloomberg.
  - Banks with regulatory capital ratios below regulatory minimums were deleted from the sample.
  - Final sample: 1,058 commercial banks (789 in the United States and 269 in Europe).
- Representativeness:
  - Final sample accounts, on average, for 81 percent of total loans from U.S. commercial banks (FDIC) and, on average, for 58 percent of total loans from European commercial banks (central banks), with country variation (e.g., 6 percent in the Slovak Republic and 84 percent in Belgium).
- Regulatory minimums noted:
  - Minimum requirement at 8 percent for total regulatory capital to total risk-weighted assets (except Cyprus at 10 percent, and United Kingdom considered equal to 9 percent following Jokipii and Milne (2008)).
  - Tier 1 capital ratio minimums: 4 percent before 2013, 4.5 percent in 2013, 5.5 percent in 2014, and 6 percent in 2015 (Basel III phase-in arrangements).
  - Ratios of Tier 1 and Core Tier 1 capital to total assets minimum at 3 percent following the “Basel III leverage ratio framework and disclosure requirements” (BIS, 2014a).
- Measurement choices:
  - Aggregate total loans used (rather than total assets) to improve comparability across accounting standards (IFRS vs. GAAP).

### IV. Stylized facts (2008–15)
- Fact 1 — Lending growth patterns:
  - U.S. banks: drop in lending-growth over 2008–11; recovery starting in 2011 stronger for commercial loans than retail-and-other loans.
    - Commercial loans growth rates: 12.8, 3.6, and 0.4 percent for 2008, 2009, 2010 respectively; recovery up to 11.8 percent in 2015.
    - Retail-and-other-loans growth rates: 14.9, 10, 3.4, and 1.6 percent for 2008–11 respectively; recovery up to 8.9 percent in 2015.
  - European banks: continuous downward trends in lending-growth over 2008–15; credit crunch since 2014.
    - Commercial loans growth rate negative since 2014, e.g., -9.6 percent (specific years indicated in source).
    - Retail-and-other-loans growth rates turning negative since 2014, e.g., -9.6 and -6.1 percent for respective years.
- Fact 2 — Capital ratios:
  - European banks increased capital ratios more significantly than U.S. banks over 2008–15, but U.S. banks have higher capital ratios, especially leverage ratios.
    - European banks:
      - Total risk-weighted capital ratio: increased from 13.5 to 17.9 percent (2008–15).
      - Tier 1 capital ratio: increased from 10.7 to 15.8 percent.
      - Tier 1 leverage ratio: increased from 6.2 to 7.3 percent.
      - Core Tier 1 leverage ratio: increased from 6.0 to 7.0 percent.
    - U.S. banks:
      - Tier 1 leverage ratio: increased from 10.2 percent in 2008 to 10.4 percent in 2015.
      - Core Tier 1 leverage ratio: increased from 9.6 percent in 2008 to 10.1 percent in 2015.
      - Total risk-weighted capital ratio: slightly increased from 15.5 percent in 2008 to 16.1 percent in 2015, peaking at 17.0 percent in 2012.
      - Tier 1 capital ratio: increased from 14.3 percent in 2008 to 14.8 percent in 2015, peaking at 15.6 percent in 2012.
- Fact 3 — Liquidity and stable funding:
  - U.S. banks hold more stable funding than European banks; European banks hold more liquid assets than U.S. banks.
  - Trends:
    - U.S. banks increased stocks of liquid assets from 29.9 percent in 2008 to 35.5 percent in 2012; then downward to 32.0 percent in 2015.
    - Available amount of stable funding (Basel III–inspired) for U.S. banks follows a downward trend from 87.8 percent in 2008 to 84.4 percent in 2015.
    - Core deposits measure for U.S. banks increased from 75.1 percent in 2008 to 81.5 percent in 2013, then declined to 78.9 percent in 2015.
  - Definitions used (as stated in source):
    - Total commercial loans: loans to commercial and industrial entities + commercial real estate loans + construction loans + loans to agriculture + loans to money market funds.
    - Total retail-and-other-loans: credit card loans + installment loans + residential mortgage loans + loans to municipalities and governments + other loans.
    - Total risk-weighted capital ratio: (Tier 1 capital + Tier 2 capital)/total risk-weighted assets.
    - T1 capital ratio: Tier 1 capital/total risk-weighted assets.
    - T1 leverage ratio: Tier 1 capital/total assets.
    - Core Tier 1 leverage ratio: Core Tier 1 capital/total assets.
    - Nonrequired amount of stable funding: nonrequired amount of stable funding (based on Basel III definition)/total assets.
    - Available amount of stable funding: available amount of stable funding (based on Basel III definition)/total assets.
    - Available amount of stable funding (including Core deposits) for U.S. banks: available amount of stable funding (Core deposits being considered as stable)/total assets.
    - Core deposit definition: sum of deposits lower than US$100,000, regardless of maturity.

### V. Literature review — capital and lending (summary)
- Two main streams in empirical literature on capital’s effect on loan supply:
  - Studies on Basel I’s introduction and macroeconomic consequences, including potential hampering of bank lending and economic growth.
  - Micro-level studies linking bank-level lending growth to initial capital ratios and other bank characteristics (example cited: Bernanke and Lown (1991) finding positive linkage between bank-lending-growth and initial capital ratios for 1990:Q2–1991:Q1, but highlighting stronger role of economic environment).
- Additional empirical findings referenced:
  - Berger and Udell (1994) showing reallocation of bank credit from loans to securities in early 1990s using virtually all U.S. banks data (1979–…).

*Source: IMF working paper (wp17245), sections I–IV, Tables and Figures list, and Appendix listings as provided in the source content.*

### 1992. One of these prescriptions was the Basel accord on risk-based capital, which mandates

### wp17245 - 1992. One of these prescriptions was the Basel accord on risk-based capital, which mandates

### Hypotheses on the role of capital
- H1.1. Under deteriorated economic conditions and the credit crunch, higher capital ratios may be associated with lower lending.
- H1.2. Higher capital allows banks to absorb greater risk and enhances their ability to originate more credit.
- Capital measures considered (Basel-based):
  - t12k_rwa: (Tier 1 capital + Tier 2 capital)/total risk-weighted assets.
  - t1k_rwa: Tier 1 capital/total risk-weighted assets.
  - t1k_ta: Tier 1 capital/total assets.
  - ct1k_ta: Core Tier 1 capital/total assets.

### Empirical evidence and hypotheses on liquidity
- H2.1. Banks with stable funding sources and buffer stocks of liquid assets can originate more credit.
- H2.2. Higher liquid assets or stable funding ratios may be associated with lower lending.
- Basel-III-inspired liquidity measures used:
  - nrasf_ta: nonrequired amount of stable funding/total assets.
  - aasf_ta: available amount of stable funding/total assets.
  - aasf_cd_ta: available amount of stable funding (Core deposits considered stable)/total assets.
- Table 2 balance sheet weighting factors (selected examples preserved exactly as in source):
  - Cash and near cash items: 1
  - Trading securities: 1
  - Consumer loans: 0.25
  - Commercial loans: 0
  - Demand & saving deposits: 0.7 (Core deposits: 1 in alternative)
  - Non-Core deposits: 0
  - Short-term borrowings: 0
  - Long-term borrowings: 1
  - Total equity: 1
- NSFR definition noted: ratio should be equal to at least 100 percent on an ongoing basis; “Available stable funding” defined as portion of capital and liabilities expected to be reliable over one year; required stable funding is a function of liquidity characteristics and residual maturities of assets and OBS exposures.

### Empirical strategy and key data definitions
- Model:
  - Annual panel data static regression model estimating lending-growth:
    - 훥LO_i,t = α_i + Σ_j β_j X_ji,t−1 + Σ_k β_k X_ki,t + ε_i,t
  - Dependent variable: lending-growth measured as year-on-year change in loans expressed in current U.S. dollars.
  - Bank-specific variables are lagged once (t-1) to mitigate endogeneity.
  - OLS panel estimator with bank cross section fixed effects is employed.
  - Time fixed effects excluded when controlling for demand effects via macro variables.
  - Standard errors robust to heteroskedasticity.
  - All variables winsorized at the 1st and 99th percentile levels.
  - Serial correlation tested via Wooldridge test; null of “no first-order autocorrelation” cannot be rejected; therefore static panel model used.
- Sample and period:
  - Pooling 1058 U.S. and European commercial banks over the period 2008–15.
  - Sample split: 789 U.S. commercial banks and 269 European commercial banks.
- Dependent variable alternates:
  - Annual growth rate of commercial loans (including commercial and industrial, commercial real estate, construction, agriculture, loans to money market funds).
  - Annual growth rate of retail-and-other-loans (credit card loans, installment loans, residential mortgage loans, loans to municipalities and governments).
- Credit risk proxy:
  - npl_tlo: ratio of nonperforming loans to total gross loans.
- Bank risk appetite proxy:
  - roa_sdroa: return on assets / three-year standard deviation of ROA.
- Funding cost:
  - cost_fund: net interest expenses/(total deposits + marketable debt securities).
- Profitability:
  - roe: net income/total equity.
- Bank size:
  - ln_ta: natural logarithm of total assets.
- Merger/acquisition dummy:
  - DUM_MA equals 1 if annual growth rate of total assets > 35 percent, 0 otherwise.
- Macroeconomic variables:
  - ngdp_gwt: annual growth rate of nominal GDP.
  - 3m_itbnk_c: change in the three-month interbank rate.
  - cb_gdp_gwt: annual growth rate of central bank assets to GDP.
- Basel III implementation index:
  - basel3: sum of three dummy variables for enforcement of risk-based capital, liquidity coverage ratios, and leverage ratios; varies between 0 and 3.

### Main empirical findings (U.S. vs Europe; capital and liquidity)
- Capital effects:
  - For U.S. banks:
    - Risk-weighted capital ratios and leverage ratios have significant and positive impacts on bank-lending-growth for both commercial and retail-and-other specifications.
    - Interpretation: capitalization encourages U.S. bank-lending-growth post-2008 as banks strengthen financial soundness and loss-absorption when expanding credit.
  - For European banks:
    - Capital ratios are not significant determinants of commercial-lending-growth.
    - Tier 1 and Tier 2 risk-weighted capital ratio and Tier 1 leverage ratios have significant and negative impacts on retail-and-other-lending-growth.
    - Interpretation: possible deleveraging and credit crunch in Europe; substitution into risk-free government securities in a very low interest-rate environment; emergence of zombie lending may distort relationships.
  - European sample capitalization context:
    - Average risk-weighted capital ratio reported as 15.9 percent.
    - Average Tier 1 leverage ratio reported as 6.8 percent.
- Liquidity effects:
  - nrasf_ta (nonrequired amount of stable funding/total assets):
    - Significant and positive impact on U.S. and European commercial-lending-growth.
    - Significant and negative impact on European retail-and-other-lending-growth.
    - Source interpretation: buffer stocks of liquid assets enable expansion in risky, illiquid commercial lending (U.S.), but enable reduction of semiliquid retail loans (Europe) amid pressures to shrink assets in low interest-rate environment.
  - aasf_ta (available amount of stable funding/total assets):
    - Not significant for U.S. and European bank-lending-growth in general.
  - aasf_cd_ta (available amount of stable funding with Core deposits for U.S. banks):
    - Significant and negative impact on U.S. retail-and-other-lending-growth.
    - Interpretation: more stable funding sources may weaken U.S. banks’ willingness to expand semiliquid retail-and-other lending; raises need to clarify regulatory definition of Core/stable deposits.
- Other significant determinants:
  - Credit risk (npl_tlo): significant and negative impact on European and U.S. commercial and retail-and-other lending-growth.
  - Bank size (ln_ta): significant and negative impact on U.S. and European commercial and retail-and-other lending-growth (small banks expand more).
  - Bank cost of funding (cost_fund): significant and positive impact on European commercial-lending-growth (inability to reduce commercial loans amid higher funding costs during 2008–15).
  - ngdp_gwt: significant and negative impact on U.S. lending-growth; significant and positive impact on European commercial-lending-growth.
  - 3m_itbnk_c: unexpected significant and positive impact on U.S. lending-growth (lower policy rates after 2008 tended to hamper U.S. lending-growth according to interpretation).
  - cb_gdp_gwt: significant and positive impact on U.S. commercial and retail-and-other lending-growth (unconventional monetary policy boosting U.S. bank credit activities).
- Numerical macro-growth context (preserved exactly from source):
  - Over 2008–15 average nominal GDP growth rate in the United States: 2.81 percent.
  - For the European countries: average nominal GDP growth rate: 1.94 percent.
  - Reported maximum nominal GDP growth rate among countries: 13.7 percent.
  - Reported minimum nominal GDP growth rate among countries: -9.6 percent.

### Heterogeneity by bank size and market access
- Small vs large banks:
  - Small bank definition: total assets below US$1 billion.
  - For small U.S. banks:
    - Capital ratios (both risk-weighted and leverage) have significant and positive impacts on lending-growth for both dependent-variable specifications.
    - Interpretation: capitalization matters more for small banks due to higher costs of raising equity and less reputational advantage.
  - For large U.S. banks:
    - Only leverage ratios have significant and positive impacts on commercial-lending-growth.
- Rationale:
  - Large banks may benefit from reputational advantages and broader access to debt and equity markets; small banks focus more on traditional intermediation and are more affected by capital constraints.

### Methodological and robustness notes
- Variables are expressed in percentages except LN_TA, DUM_MA, and BASEL3 in correlation tables.
- Correlation tables provided for U.S. and European samples (selected coefficients noted in discussion, e.g., t12k_rwa and t1k_rwa correlation values reported in tables).
- Winsorization applied at 1st and 99th percentiles to reduce outlier effects.
- Haussmann test used to justify fixed-effects over random-effects.
- Wooldridge test for serial correlation: null of no first-order autocorrelation cannot be rejected.
- Regressions run separately for U.S. and European banks and for large and small banks to capture heterogeneity.
- Basel III degree-of-implementation index (basel3) varies between 0 and 3 and is included to capture regulatory heterogeneity across countries and time.

*Source: wp17245 - 1992. One of these prescriptions was the Basel accord on risk-based capital, which mandates (IMF working paper content unit).*

### 63.6 percent for large European banks, and 69.0 percent for small European banks. By type of loans granted, the

### wp17245 - 63.6 percent for large European banks, and 69.0 percent for small European banks. By type of loans granted, the

### Descriptive statistics and loan shares
- Average share of commercial loans in total assets on the whole sample: 41.4 percent.
- Commercial loans by subgroup:
  - Large U.S. banks: 45.4 percent.
  - Small U.S. banks: 43.6 percent.
  - Large European banks: 31.6 percent.
  - Small European banks: 31.9 percent.
- Average share of retail-and-other-loans in total assets on the whole sample: 25.3 percent.
  - Large U.S. banks: 21.2 percent.
  - Small U.S. banks: 24.7 percent.
  - Large European banks: 31.2 percent.
  - Small European banks: 37.5 percent.
- Average share of deposits in total assets on the whole sample: 74.4 percent.
  - Large U.S. banks: 77.4 percent.
  - Small U.S. banks: 81.5 percent.
  - Large European banks: 53.0 percent.
  - Small European banks: 70.2 percent.
- Bank performance and asset quality (descriptive):
  - Small U.S. banks ROE: 3.0 percent.
  - Small U.S. banks nonperforming loans: 3.1 percent.
  - Large U.S. banks ROE: 5.9 percent.
  - Large U.S. banks nonperforming loans: 2.6 percent.

### Impacts of capital ratios on lending-growth
- Large U.S. banks:
  - Capital ratios have significant and negative impacts on large bank-retail-and-other-lending-growth (context: deleveraging and credit crunch in Europe post-2008).  
  - Large U.S. banks are more involved in securitization lending than small banks; investing in illiquid commercial loans strengthens financial soundness and loss-absorption capacity.
- Large European banks:
  - Capital ratios—both risk-weighted and simple leverage—have significant and negative impacts on large bank-retail-and-other-lending-growth.
  - Interpretation: increases in capital may act as a regulatory tax higher on assets assigned higher risk weights or when expanding balance sheets via credit activities.
  - Large European banks are more involved in trading than in credit activities—especially retail lending—compared to small European banks.
  - Enforcement of more stringent capital adequacy rules is expected to encourage large banks to substitute out of retail-and-other-loan assets into risk-free, more liquid government bond securities.

### Impacts of liquidity ratios on lending-growth
- Ratio of the nonrequired amount of stable funding to total assets (nraf_ta):
  - Significant and positive impact on commercial-lending-growth for U.S. banks, regardless of size.
    - Interpretation: when expanding commercial lending, U.S. banks hold buffer stocks of liquid assets to face unexpected liquidity disruptions.
  - Significant and positive impact on commercial-lending-growth for large European banks.
    - Interpretation: large European banks build up marketable liquid asset buffers to enhance risk absorption and avoid cash shortfalls when holding riskier, illiquid loans.
  - Significant and negative impact on large European bank-retail-and-other-lending-growth.
    - Interpretation: large European banks use marketable liquid asset buffers to reduce retail-and-other-lending-activities amid pressures to shrink assets, especially in a very low interest-rate environment (post-2008).
  - Note: nraf_ta has a significant and negative impact on small U.S. bank-retail-and-other-lending-growth only in models that include risk-weighted capital ratios; this impact is weakly significant at the 10 or 5 percent levels and is not emphasized.
- Ratio of the available amount of stable funding to total assets (aasf_ta):
  - Not significant in the determination of U.S. and European bank-lending-growth, regardless of size (consistent with previous results).

### Core-deposit-focused funding effects
- Ratio of the available amount of stable funding that focuses more closely on Core deposits (aasf focusing on Core deposits):
  - Large U.S. banks:
    - Significant and negative impact on large U.S. bank-retail-lending-growth.
    - Interpretation: more stable funding sources weaken large U.S. banks’ willingness to expand semiliquid retail-and-other-lending activities; large banks may underestimate funding liquidity risks due to broader access to external funding and allocate less stable funding when investing in semiliquid retail-and-other-loans.
  - Small U.S. banks:
    - Significant and positive impact on small U.S. bank-commercial-lending-growth.
    - Interpretation: small U.S. banks rely on stable Core-deposit funding sources to expand risky and illiquid commercial-lending activities; they hold strong security buffers of stable funding when expanding illiquid credit activities due to difficulty raising external market funding.

### Robustness checks and methodological notes
- Endogeneity concerns:
  - Bank-level explanatory variables presumed endogenous are replaced by one-year lagged values; model estimated using generalized method of moments (GMM) as a robustness check.
  - One-year lagged values are not weak instruments; more lags are weak instruments and thus excluded.
  - Cross-section fixed effects included; time fixed effects excluded when controlling for demand effects via country-level time series.
  - Most results consistent with OLS estimates; some effects weakened:
    - Capital ratios become not significant for large U.S. banks in determination of commercial-lending-growth.
    - Capital ratios become not significant for small U.S. banks in determination of retail-and-other-lending-growth.
    - Negative impact of capital ratio weakened for large European banks in determination of retail-and-other-lending-growth.
    - Ratio of available amount of stable funding focusing on Core deposits for U.S. banks becomes significant with a negative impact on large U.S. bank-commercial-lending-growth—consistent with interpretation that large U.S. banks might underestimate funding liquidity risks and allocate less stable funding when investing in illiquid commercial loans.
- Alternative dependent variable:
  - Using ratio of retail loans to total assets as dependent variable produces main conclusions consistent with those obtained when including other loans in addition to retail lending.
- Inclusion of banks below regulatory capital minima:
  - Including banks with regulatory capital ratios below minimum requirements increases observations; main conclusions remain consistent.
- Alternative weights for Basel III–inspired liquidity ratios:
  - Consumer loan weight alternatives: 0.15 (nradf015_ta) and 0.5 (nradf05_ta) versus baseline 0.25.
  - Demand and saving deposit weight alternatives: 0.5 (aasf05_ta) and 0.9 (aasf09_ta) versus baseline 0.7.
  - Main conclusions consistent across these alternative weight specifications.
- Systemically important and size-based subgroup checks:
  - Subsamples: (i) very large (significant) banks with total assets above US$50 billion (all G-SIFIs included in sample remain in this group), and (ii) other large banks with total assets below US$50 billion and above US$1 billion.
  - Results consistent with pooling all large banks above US$1 billion; ratio of available amount of stable funding focusing on Core deposits for U.S. banks has significant negative impact on bank-retail-lending-growth only for other large U.S. banks (not very large).
  - Terminology note: Federal Reserve qualifies a bank as significant if it holds US$50 billion or more in total consolidated assets (FED, 2011).
- European bank size alternative:
  - Sample of European banks includes mainly large banks with average total assets of US$160 billion and median total assets of US$20 billion.
  - Alternative definition: European bank considered small if total assets below US$20 billion (median); regression results reported in Table A1.4 (main conclusions consistent).

*Source: wp17245 (excerpt).*

### conclusions are consistent with those previously obtained when including an alternative

### wp17245 - conclusions are consistent with those previously obtained when including an alternative

### Main results on capital measures and lending
- Study period: 2008–15.
- Small U.S. banks strengthen their capital base and loss absorption capacity when expanding both commercial and retail-and-other-credit-activities.
- Large U.S. banks only strengthen their leverage ratios when granting riskier, illiquid commercial loans.
- Capital ratios have significant and negative impacts on bank-retail-and-other-lending-growth for large European banks.
- In Europe’s deleveraging and credit crunch over the post-2008 period, more stringent capital adequacy rules encourage substitution out of retail-and-other-loan assets and into risk-free, more liquid government bond securities.

### Liquidity indicators and perverse effects on lending
- The ratio of the nonrequired amount of stable funding to total assets (nrasf_ta) has a significant, positive impact on commercial-lending-growth for U.S. banks, regardless of size.
- For large European banks, nrasf_ta has a positive impact on commercial-lending-growth but a negative effect on retail-lending-growth — consistent with an inability to reduce commercial loans while curtailing retail-and-other-lending when holding liquid asset buffers.
- The ratio of the available amount of stable funding to total assets (aasf_ta) is not significant in determining European bank-lending-growth regardless of size.
- The ratio of the available amount of stable funding that focuses on Core deposits for U.S. banks (aasf_cd_ta) has:
  - A significant and negative impact on retail-lending-growth for large U.S. banks.
  - A significant and positive impact on commercial-lending-growth for small U.S. banks.
- Interpretation: U.S. banks likely hold buffer stocks of liquid assets to pre-empt unexpected liquidity disruptions when expanding risky, illiquid credit activities; funding structure matters for allocation between commercial and retail-and-other lending.

### Bank-size heterogeneity (large vs small; U.S. vs European)
- Bank-size threshold used in some analyses: US$1 billion (a bank is considered large if its total assets are above US$1 billion; small if below US$1 billion).
- Small European banks: inability to reduce commercial loans amid pressures to shrink assets when holding buffer stocks of liquid assets; they display cautious behavior and hold strong security buffers of liquid assets when unable to reduce exposure to riskier, illiquid loans.
- Large European banks: capital ratios negatively affect retail-and-other-lending-growth; funding composition determines whether banks curtail retail activities while sustaining commercial loans.
- Small U.S. banks expand risky and illiquid commercial-lending-activities when relying more on stable Core deposit funding sources; large U.S. banks may benefit from broader access to external funding to allocate less stable funding when expanding semiliquid retail-and-other lending.

### Empirical evidence (selected model findings and diagnostics)
- Regression framework: OLS on unbalanced panels of commercial banks over 2008–15; bank-specific explanatory variables replaced by one-year lagged values; standard errors robust to heteroskedasticity.
- Table-level diagnostic indicators (examples from reported regressions):
  - Table 5 (U.S. banks, various specifications): r2 values around 0.221 to 0.225 for commercial-loans regressions; F-statistics range e.g., 68.020, 78.890; N values reported (examples) 5130, 5669.
  - Table 6 (European banks): r2 around 0.194–0.213; F-statistics around 13.94–14.86; N 1595 (various specifications 1595, 1536, 1474).
  - Table 7 (Large U.S. banks): r2 for commercial loans 0.262–0.321; for retail-and-other loans 0.087–0.104; N 1810, 1974, 1949.
  - Table 8 (Small U.S. banks): r2 for commercial loans 0.262–0.308; for retail-and-other loans 0.084–0.121; N 3320, 3695, 3669.
  - Table 9 (Large vs Small European banks): N 1507 and N 88 in reported subsamples; r2 vary widely across small-bank specifications (e.g., 0.742 down to 0.500).
- Reported statistical significance conventions: * 10 percent; ** 5 percent; *** 1 percent. Student-t statistics are reported in parentheses in tables.

### Policy implications and recommendations
- The effect of new capital and liquidity regulatory frameworks may differ depending on economic context and a bank’s ability to originate credit; they may be ineffective or even detrimental in some contexts.
- Implementing capital and liquidity regulatory frameworks uniformly across all banks could have counterproductive effects on post-2008 bank-lending-growth.
- Regulators need clearer definitions and measurements under a global regulatory framework — specifically, determination of what types of liquid liabilities should be considered stable for a deeper regulatory definition of the notion of stable funding.
- Key recommendation: Regulators should consider heterogeneous banks’ characteristics and behaviors when designing and enforcing regulations to promote stronger financial stability and to strengthen banks’ core functions as credit providers.

*Source: wp17245 - conclusions are consistent with those previously obtained when including an alternative*

### REFERENCES

### wp17245 - REFERENCES

### Bibliographic references (selection)
- A comprehensive list of cited works covering monetary policy transmission, bank-lending behavior, bank liquidity and capital, Basel III, and empirical methods. Notable cited authors and institutions include: Abdkarim, Acharya, Aiyar, Alfaro, Allen, Atkinson, Baltagi, Bank of International Settlements, Beatty, Ben Naceur, Berger, Bernanke, Berrospide, Borio, Bridges, Carlson, Chami, Cornett, Dagher, De Young, Ehrmann, Elshahat, European Central Bank, Financial Stability Board, Gambacorta, Hall, Ivashina, Jokipii, Kashyap, Keeton, Kishan, Košak, Kupiec, Labonne, Ladime, Laidroo, Lepetit, Olszak, Pana, Panetta, Peek, Peersman, Pruteanu-Podpiera, Saunders, Setiyono, Stein, Stiglitz, Talavera, Uchida, Woolridge, Wu.
- Topics covered in the references include: bank-lending channel, liquidity creation, capital requirements, non-performing loans, bank supervision, generalized method of moments, panel data econometrics, Basel III frameworks (leverage ratio, net stable funding ratio, phase-in arrangements), and country- and region-specific studies (U.S., Europe, Chile, Ghana, Malaysia, Mexico, Ukraine, Japan, CEE).

### Appendix — Empirical tables: overview and key reported estimates
- The Appendix contains empirical tables estimating equation (1) for unbalanced panels of commercial banks over the period 2008–15. Methods used include GMM and standard OLS. Bank-specific explanatory variables and several macroeconomic indicators are replaced by their one-year lagged values. Standard errors are robust to heteroskedasticity. Significance notation: * 10 percent, ** 5 percent, *** 1 percent. Definitions of variables are referenced to Table 4 in the source.

- Table A1.1a. The Determinants of Large U.S. Banks’ Lending Using Generalized Method of Moments During 2008–15
  - Sample sizes: N 1810, 1974, 1949 across specifications.
  - Commercial loans — selected coefficient estimates (standard errors in parentheses):
    - nrasf_ta: 0.273 (0.214); 0.275 (0.210); 0.369* (0.192); 0.432** (0.191); 0.495*** (0.177); 0.478*** (0.175); 0.576*** (0.170); 0.596*** (0.174).
    - aasf_ta: -0.469** (0.228); -0.454** (0.224); -0.281 (0.219); -0.257 (0.212).
    - aasf_cd_ta: -0.141* (0.0729); -0.137* (0.0726); -0.124 (0.0773); -0.131* (0.0777).
  - Retail and other loans — selected coefficient estimates:
    - t12k_rwa: 0.813** (0.407); 0.895** (0.425).
    - t1k_rwa: 0.818* (0.418); 0.855** (0.434).
    - aasf_cd_ta: -0.314** (0.135); -0.305** (0.134); -0.304** (0.124); -0.291** (0.121).

- Table A1.1b. The Determinants of Small U.S. Banks’ Lending Using Generalized Method of Moments During 2008–15
  - Sample sizes: N 3320, 3695, 3669 across specifications.
  - Commercial loans — selected coefficient estimates:
    - t1k_ta: 1.508*** (0.424); 1.169*** (0.420).
    - ct1k_ta: 1.446*** (0.428); 1.138*** (0.426).
    - nrasf_ta: 0.678*** (0.157); 0.699*** (0.151); 0.843*** (0.142); 0.852*** (0.147); 0.555*** (0.154); 0.603*** (0.150); 0.796*** (0.144); 0.841*** (0.149).
    - aasf_cd_ta: 0.0951* (0.0490); 0.0833* (0.0475); 0.0777* (0.0462); 0.0777* (0.0468).
  - Retail and other loans — selected coefficient estimates (insignificant or smaller magnitudes noted):
    - nrasf_ta: 0.185 (0.251); 0.149 (0.242); 0.248 (0.218); 0.262 (0.223); 0.217 (0.257); 0.177 (0.250); 0.216 (0.224); 0.211 (0.228).
    - aasf_cd_ta: -0.025 (0.0813); -0.023 (0.0814); -0.008 (0.0702); -0.01 (0.0698).

- Table A1.1c. The Determinants of European Banks’ Lending Using Generalized Method of Moments for Large vs. Small European Banks During 2008–15
  - Large banks — sample sizes: N 1507, 1448, 1386 across specifications.
    - Commercial loans — selected coefficient estimates:
      - t12k_rwa: -0.971** (0.482); -0.737** (0.347).
      - nrasf_ta: 0.953*** (0.304); 0.976*** (0.324); 1.348*** (0.350); 1.299*** (0.351).
    - Retail and other loans — selected coefficient estimates:
      - nrasf_ta: -0.333* (0.298); -0.368* (0.307); -0.602* (0.343); -0.661* (0.355).
  - Small banks — sample size: N 88 across specifications.
    - Coefficients generally estimated with larger standard errors; selected estimates include:
      - t12k_rwa: -1.886 (1.587); -0.81 (1.363).
      - t1k_rwa: -2.069 (1.672); -1.323 (1.186).
      - nrasf_ta: -0.682 (0.645); -0.698 (0.619); -0.822 (1.061); -0.774 (0.917).

- Table A1.2a. The Determinants of Large U.S. Banks’ Lending Using Alternative Weights for Liquidity Ratios Inspired by Basel III During 2008–15
  - Estimation method: standard OLS for an unbalanced panel of U.S. commercial banks over 2008–15. Student-t statistics reported in parentheses. Weights in liquidity ratios are varied (e.g., nrasf015_ta, nrasf05_ta, nrasf09_ta).
  - Sample sizes across columns: N 1810, 1974, 1949.
  - Commercial loans — selected coefficient estimates:
    - t12k_rwa: 0.183 (0.99); 0.181 (0.99); 0.184 (1.00).
    - t1k_rwa: 0.146 (0.76); 0.145 (0.75); 0.146 (0.76).
    - t1k_ta: 0.454 (1.63); 0.455 (1.63); 0.45 (1.61).
    - ct1k_ta: 0.686** (2.35); 0.686** (2.35); 0.683** (2.33).
    - nrasf015_ta: 0.522*** (5.27); 0.520*** (5.29); 0.528*** (5.30); 0.524*** (5.30); 0.522*** (5.32); 0.530*** (5.32); 0.591*** (6.21); 0.588*** (6.23); 0.595*** (6.24); 0.596*** (6.21); 0.594*** (6.22); 0.600*** (6.23).
    - aasf_ta: -0.106 (-1.07); -0.111 (-1.11); -0.086 (-0.91).
    - aasf05_ta: -0.089 (-1.36); -0.091 (-1.39); -0.077 (-1.25); -0.076 (-1.25).
    - aasf09_ta: -0.047 (-0.25); -0.06 (-0.32); -0.041 (-0.23); -0.045 (-0.24).
    - Reported r2 across specifications: 0.297; 0.295; 0.300; 0.301; 0.299; 0.263; 0.264; 0.262; 0.252; 0.254; 0.252 (varies by column).
  - Retail and other loans — selected coefficient estimates:
    - t12k_rwa: 0.272 (0.86); 0.276 (0.87).
    - t1k_rwa: 0.405 (1.30); 0.404 (1.29); 0.411 (1.32).
    - t1k_ta: 0.878* (1.90); 0.870* (1.88); 0.909** (1.97).
    - nrasf015_ta and variants: coefficients generally smaller and less statistically significant for retail loans (examples: 0.153 (1.10); 0.157 (1.13); 0.139 (1.00); 0.134 (0.98)).
    - aasf09_ta reported negative and sometimes significant: -0.444* (-1.74); -0.446* (-1.75); -0.550** (-2.31); -0.552** (-2.28).
    - Reported r2 values for retail specifications: 0.087; 0.089; 0.096; 0.093; 0.092 (varies by column).

### Appendix notes (methodology and sample definitions)
- Estimation approaches: GMM for panel specifications (Tables A1.1a–c); standard OLS for robustness/alternative weighting (Table A1.2a).
- Time period covered: 2008–15.
- Lag structure: bank-specific explanatory variables and several macroeconomic indicators replaced by their one-year lagged values.
- Bank size cutoff: a bank is considered large if its total assets are above US$1 billion; small if its total assets are below US$1 billion.
- Standard errors: robust to heteroskedasticity. Student-t statistics or standard errors are reported in parentheses as indicated in each table.

*Italicized footnote in source: Table notes and variable definitions are provided in Table 4 of the source document.*

### 0.15 and 0.5. The 0.7 weight for demand and saving deposits is changed to 0.5 and 0.9. A bank is considered large if its

### 0.15 and 0.5. The 0.7 weight for demand and saving deposits is changed to 0.5 and 0.9. A bank is considered large if its total assets are above US$1 billion.

### Regression results — Commercial loans (selected coefficients and statistics)
- t12k_rwa: 0.664*** (3.32); 0.664*** (3.32); 0.658*** (3.27)
- t1k_rwa: 0.638*** (3.07); 0.638*** (3.07); 0.632*** (3.02)
- t1k_ta: 1.058*** (4.01); 1.058*** (4.01); 1.049*** (3.97)
- ct1k_ta: 1.063*** (4.04); 1.062*** (4.04); 1.051*** (3.98)
- nrasf015_ta: 0.582*** (7.47); 0.582*** (7.49); 0.588*** (7.49); 0.583*** (7.48); 0.583*** (7.49); 0.588*** (7.48); 0.697*** (9.35); 0.697*** (9.35); 0.701*** (9.40); 0.707*** (9.32); 0.706*** (9.32); 0.711*** (9.37)
- aasf_ta: -0.15 (-1.52); -0.157 (-1.61); -0.145 (-1.62); -0.158* (-1.77)
- aasf05_ta: -0.0976* (-1.67); -0.102* (-1.75); -0.0947* (-1.77); -0.103* (-1.91)
- aasf09_ta: -0.18 (-0.75); -0.197 (-0.82); -0.17 (-0.80); -0.197 (-0.93)
- r2: 0.247; 0.248; 0.246; 0.251; 0.252; 0.250; 0.280; 0.281; 0.278; 0.273; 0.273; 0.271
- N: 3320; 3320; 3695; 3669; 3320; 3320; 3695; 3669; 3320; 3320; 3695; 3669

### Regression results — Commercial loans (alternate set / smaller-sample coefficients)
- t12k_rwa: 0.485** (2.43); 0.485** (2.43); 0.480** (2.40)
- t1k_rwa: 0.477** (2.30); 0.477** (2.31); 0.472** (2.27)
- t1k_ta: 1.188*** (4.68); 1.188*** (4.68); 1.182*** (4.65)
- ct1k_ta: 1.205*** (4.78); 1.204*** (4.78); 1.197*** (4.73)
- nrasf05_ta: 0.949*** (10.66); 0.949*** (10.67); 0.953*** (10.68); 0.947*** (10.42); 0.947*** (10.43); 0.951*** (10.42); 1.014*** (12.86); 1.014*** (12.86); 1.017*** (12.88); 1.029*** (12.93); 1.029*** (12.93); 1.032*** (12.95)
- aasf_ta: -0.071 (-0.73); -0.082 (-0.86); -0.087 (-1.00); -0.101 (-1.16)
- aasf05_ta: -0.048 (-0.85); -0.055 (-0.97); -0.057 (-1.10); -0.065 (-1.26)
- aasf09_ta: -0.039 (-0.16); -0.063 (-0.26); -0.087 (-0.41); -0.117 (-0.55)
- r2: 0.307; 0.308; 0.307; 0.313; 0.313; 0.311; 0.354; 0.354; 0.353; 0.348; 0.348; 0.346
- N: 3320; 3320; 3695; 3669; 3320; 3320; 3695; 3669; 3320; 3320; 3695; 3669

### Regression results — Retail and other loans (selected coefficients and statistics)
- t12k_rwa: 0.689** (2.11); 0.692** (2.12); 0.682** (2.09)
- t1k_rwa: 0.715** (2.30); 0.718** (2.31); 0.709** (2.28)
- t1k_ta: 0.916** (2.23); 0.922** (2.24); 0.908** (2.20)
- ct1k_ta: 0.724* (1.73); 0.731* (1.75); 0.712* (1.70)
- nrasf015_ta: -0.059 (-0.45); -0.061 (-0.47); -0.054 (-0.41); -0.059 (-0.46); -0.061 (-0.48); -0.054 (-0.42); 0.096 (0.89); 0.094 (0.88); 0.099 (0.92); 0.104 (0.96); 0.102 (0.94); 0.106 (0.99)
- aasf_ta: 0.137 (0.98); 0.138 (0.99); 0.259* (1.86); 0.266* (1.90)
- aasf05_ta: 0.068 (0.76); 0.068 (0.78); 0.142 (1.64); 0.147* (1.68)
- aasf09_ta: 0.458 (1.43); 0.453 (1.42); 0.682** (2.20); 0.688** (2.21)
- r2: 0.085; 0.085; 0.084; 0.093; 0.094; 0.092; 0.112; 0.113; 0.111; 0.099; 0.100; 0.098
- N: 3320; 3320; 3695; 3669; 3320; 3320; 3695; 3669; 3320; 3320; 3695; 3669

### Additional coefficient highlights (selected)
- t12k_rwa (other rows): 1.080*** (3.28); 1.082*** (3.28); 1.073*** (3.25)
- t1k_rwa (other rows): 1.079*** (3.44); 1.081*** (3.45); 1.072*** (3.42)
- t1k_ta (other rows): 0.699* (1.70); 0.703* (1.71); 0.690* (1.68)
- ct1k_ta (other rows): 0.489 (1.18); 0.496 (1.20); 0.477 (1.15)
- nrasf05_ta (other rows): -0.751*** (-4.68); -0.752*** (-4.70); -0.745*** (-4.64); -0.741*** (-4.70); -0.742*** (-4.71); -0.736*** (-4.65); -0.530*** (-3.87); -0.531*** (-3.88); -0.527*** (-3.85); -0.536*** (-3.88); -0.537*** (-3.90); -0.533*** (-3.87)
- aasf_ta (other rows): 0.016 (0.11); 0.024 (0.17); 0.171 (1.25); 0.178 (1.29)
- aasf05_ta (other rows): -0.005 (-0.06); 0.003 (0.00); 0.089 (1.04); 0.094 (1.08)
- aasf09_ta (other rows): 0.215 (0.68); 0.221 (0.70); 0.512* (1.69); 0.521* (1.71)
- r2 (other rows): 0.183; 0.183; 0.180; 0.186; 0.187; 0.184; 0.158; 0.159; 0.156; 0.147; 0.148; 0.145
- N (other rows): 3320; 3320; 3695; 3669; 3320; 3320; 3695; 3669; 3320; 3320; 3695; 3669

### Notes on estimation and specification
- Estimation: standard OLS for an unbalanced panel of U.S. commercial banks over the period 2008–15.
- All bank-specific explanatory variables, and several macroeconomic indicators, are replaced by their one-year lagged values.
- Standard errors are robust to heteroskedasticity.
- Significance notation: * 10 percent; ** 5 percent; *** 1 percent. Student-t statistics reported in parentheses.
- Weights are changed in the liquidity ratios inspired by Basel III. The 0.25 weight for consumer loans is changed to 0.15 and 0.5. The 0.7 weight for demand and saving deposits is changed to 0.5 and 0.9. A bank is considered large if its total assets are above US$1 billion.

*Source: wp17245*

### 0.15 and 0.5. The 0.7 weight for demand and saving deposits is changed to 0.5 and 0.9. A bank is considered small if its

### wp17245 - 0.15 and 0.5. The 0.7 weight for demand and saving deposits is changed to 0.5 and 0.9. A bank is considered small if its total assets are below US$1 billion.

### Liquidity-weight sensitivity in regressions (Basel III–inspired alternative weights)
- Weights adjusted in liquidity ratios: the 0.25 weight for consumer loans is changed to 0.15 and 0.5; the 0.7 weight for demand and saving deposits is changed to 0.5 and 0.9.
- Alternative weight specifications are used across tables A1.2c, A1.2d, and other robustness checks to assess sensitivity of lending determinants to liquidity-ratio construction.

### Determinants of large European banks’ lending (Table A1.2c; 2008–15; large = total assets above US$1 billion)
- Commercial loans regressions:
  - nrasf015_ta coefficients: 0.898*** (4.90), 0.898*** (4.91), 0.897*** (4.89), 0.893*** (4.66), 0.893*** (4.67), 0.892*** (4.65), 0.863*** (4.40), 0.861*** (4.40), 0.864*** (4.40), 0.835*** (4.22), 0.834*** (4.23), 0.835*** (4.22).
  - aasf_ta and its variants show negative coefficients with occasional statistical significance: aasf_ta -0.179* (-1.65) in one model; aasf05_ta and aasf09_ta negative up to -0.210* (-1.86) and -0.182 (-1.45) in variants.
  - R-squared values across models: 0.173, 0.174, 0.173, 0.175, 0.175, 0.174, 0.173, 0.173, 0.172, 0.183, 0.184, 0.182.
  - Sample sizes N: 1507, 1507, 1448, 1386 (varies by specification).
- Other reported coefficients (t12k_rwa, t1k_rwa, t1k_ta, ct1k_ta) are generally small in magnitude and frequently statistically insignificant in the commercial-loans specifications shown.
- Retail and other loans regressions:
  - nrasf05_ta coefficients: 1.533*** (6.19), 1.532*** (6.19), 1.533*** (6.19), 1.529*** (5.95), 1.528*** (5.95), 1.529*** (5.95), 1.545*** (5.85), 1.542*** (5.84), 1.547*** (5.87), 1.491*** (5.62), 1.490*** (5.62), 1.492*** (5.64).
  - aasf_ta and aasf variants negative and sometimes significant: aasf_ta -0.227** (-2.01) in one specification; aasf09_ta -0.263** (-2.26) in another.
  - R-squared values higher than commercial loans models: 0.262–0.286 range.
  - N: same pattern (1507/1448/1386).
- Summary note: Standard errors robust to heteroskedasticity; explanatory variables lagged one year.

### Determinants of small European banks’ lending (Table A1.2d; 2008–15; small = total assets below US$1 billion)
- Commercial loans regressions (N = 88 for all models shown):
  - nrasf015_ta coefficients generally negative and not statistically significant across specifications (examples: -0.73 (-1.08), -0.769 (-1.12), -0.653 (-1.00)).
  - aasf_ta and aasf variants are negative and larger in magnitude relative to large-bank estimates (examples: aasf_ta -0.959 (-1.21), -1.164 (-1.39), -1.066 (-1.29)).
  - R-squared values vary widely across specifications: 0.758, 0.829, 0.726, 0.666, 0.745, 0.664, 0.534, 0.616, 0.520, 0.538, 0.608, 0.527.
- Retail and other loans regressions (N = 88):
  - t12k_rwa, t1k_rwa, t1k_ta, ct1k_ta estimates show large point estimates with low significance (e.g., t1k_ta 2.183 (-1.01), ct1k_ta 2.249 (-0.87) in one specification).
  - aasf_ta and variants remain negative (e.g., aasf_ta -0.875 (-1.10), -1.011 (-1.26)).
  - R-squared values around 0.414–0.441 in reported blocks and 0.417–0.432 in continued blocks.

### U.S. banks — Very large and other large (Tables A1.3a, A1.3b)
- Very Large U.S. banks (Table A1.3a; very large defined as total assets below US$50 billion; N values: 133, 133, 149, 145, etc.):
  - Commercial loans: nrasf_ta positive and significant in multiple specifications (e.g., 0.399** (2.54), 0.546*** (3.45), 0.538** (2.20), 0.669*** (4.22)).
  - aasf_ta positive and marginally significant in some specifications: 0.317* (1.88).
  - R-squared for commercial loans: 0.532–0.550; Retail loans R-squared: 0.219–0.381.
- Other Large U.S. banks (Table A1.3b; total assets between US$50 billion and US$1 billion; N: 1677, 1825, 1804):
  - Commercial loans: nrasf_ta strongly positive and highly significant (e.g., 0.632*** (6.04), 0.703*** (7.04)).
  - Retail loans: generally low and often insignificant relationships for nrasf_ta; aasf_cd_ta displays negative and sometimes significant coefficients for retail lending (e.g., -0.178* (-1.91), -0.169** (-1.99)).

### European banks — Very large vs. other large (Table A1.3c) and alternative size separations (Table A1.4)
- Very large European banks (very large defined as total assets below US$50 billion):
  - Commercial loans: nrasf_ta positive and significant (0.884*** (3.55), 0.936*** (3.89)).
  - Retail and other loans: nrasf_ta coefficients vary with some negative and significant estimates (e.g., -0.356* (-1.89), -0.410** (-2.18)).
  - aasf_ta negative and sometimes significant for retail lending (e.g., -0.227* (-1.97), -0.285** (-2.20)).
- Other large European banks (total assets between US$50 billion and US$1 billion):
  - Commercial loans: nrasf_ta positive and significant (1.305*** (4.10), 1.317*** (3.94)).
  - Retail loans: nrasf_ta negative and sometimes significant (e.g., -0.523* (-1.79), -0.489* (-1.65)); aasf_ta negative and significant for commercial lending in some specs (e.g., -0.307** (-2.01)).
- Alternative size thresholds (Table A1.4; large defined as total assets above US$20 billion):
  - Large banks (N: 792/743/683 depending on specification):
    - Commercial loans: nrasf_ta positive and significant (1.208*** (4.87), 1.201*** (4.75), 1.207*** (3.91), 1.100*** (3.78)).
    - Retail and other loans: nrasf_ta negative and significant (e.g., -0.495** (-2.55), -0.655*** (-3.64)).
    - R-squared for large-bank commercial models: 0.190–0.209; F-statistics reported (e.g., F 7.79, p 0.00).
  - Small banks (N: 803/793/791):
    - Commercial loans: nrasf_ta positive and significant in some specs (0.942*** (2.80), 0.949*** (2.72)); aasf_ta negative and occasionally significant (e.g., -0.340* (-1.88)).
    - R-squared for small-bank models: around 0.175–0.180 for commercial loans; F-statistics reported (e.g., F 8.72, p 0.00).

### Methodological and notation notes (as reported)
- Estimation: equation (1) using standard OLS for unbalanced panels over 2008–15.
- All bank-specific explanatory variables and several macroeconomic indicators replaced by one-year lagged values.
- Standard errors robust to heteroskedasticity.
- Significance notation: * at 10 percent; ** at 5 percent; *** at 1 percent. Student-t statistics reported in parentheses.
- Definitions used in captions: small banks = total assets below US$1 billion; large banks in some tables = above US$1 billion or above US$20 billion; very large = below US$50 billion; other large between US$50 billion and US$1 billion.

*Source: wp17245 (excerpted tables and notes from the supplied PDF content).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp17245.pdf_