## 1. Introduction

## Source details

**Canonical URL:** [1. Introduction](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020103-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020103-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020103-print-pdf.pdf.json)

---

### Overview and research question
- Financial crises impose large costs by requiring public funds to revive financial institutions and by destroying the finance-growth nexus.
- Main question: Do tighter bank regulations (higher liquidity/capital levels as in Basel III) affect sectoral economic activity, and does this effect differ between normal times and crisis periods?
- Identification strategy: interact a country’s bank-system soundness (liquidity and/or capital levels) with an industry’s external finance dependence to test whether sectors more reliant on external finance perform disproportionately better or worse in countries with higher bank liquidity/capital.

### Empirical approach and scope
- Bank-level dataset for 50 countries over the period 2000–10 to construct Basel III-style liquidity and capital indicators.
- Merged with data on industry activity for 28 manufacturing sectors around the period of the global financial crisis.
- Two sectoral outcomes:
  - Investment rate = gross fixed capital formation divided by output.
  - Growth in the number of establishments = proxy for entrepreneurship and new business creation.

### Basel III-style bank indicators
- NSFR (Net Stable Funding Ratio) as proxy for liquidity.
- Capital_Total = total regulatory capital ratio (Tier 1 + Tier 2 to risk-weighted assets).
- Capital_Tier1 = Tier 1 capital ratio to risk-weighted assets.
- PCA_ALL = first principal component of NSFR, Capital_Total, Capital_Tier1.
- Basel III calibration references and assumed minima: at least 10.5% for total regulatory capital and 8.5% for Tier 1.

---

### Theoretical expectations and channels
- Two competing effects of higher liquidity/capital:
  - Positive channel: higher capital increases banks’ risk-bearing capacity, potentially expanding credit supply and supporting investment and firm creation.
  - Negative channel: tighter requirements could lead banks to hoard liquidity or reduce lending to non-financial firms, lowering activity in the short run.
- Net effect may depend on macro-financial conditions:
  - During normal times, costs of tighter regulation may lower activity.
  - During crises, safer banks reduce market frictions, lower the cost of external finance, and can encourage investment and new firm formation.
- Sectors with higher external finance dependence are expected to be more sensitive to bank liquidity/capital levels.

### 2. Model, data, and identification

### Baseline empirical specification
- Framework: y_{i,c,t} = ϑ + ϕ1·Share_{i,c,t−1} + ϕ2·R_{c,t} + ϕ3·R_{c,t}×ExtDep_i + ϕ4·FinDev_{c,t} + ϕ5·FinDev_{c,t}×ExtDep_i + ε_{i,c,t}
  - y_{i,c,t}: investment rate or establishment growth.
  - ExtDep: external finance dependence at the sectoral level (US-based Rajan and Zingales (1998) benchmark).
  - R_{c,t}: country average bank liquidity or capital ratio.
  - Main parameter of interest: ϕ3 on R_{c,t}×ExtDep_i.
- Fixed effects: sector, country, and year fixed effects.
- Estimation: OLS with standard errors clustered by industry; robustness checks include country clustering and two-way clustering.

### Data sources and construction
- Bankscope bank-level data: 1,857 banks from 50 developed and emerging economies, 2000–2010. Country-level bank measures = unweighted averages of bank-level ratios.
- Industry data: UNIDO Industrial Statistics Database; 73 ISIC Rev.3 groups regrouped into 28 industries (ISIC Rev.2).
- External finance dependence (ExtDep) from Rajan and Zingales (1998), using US firm-level data.
- Financial development (FinDev) and other country variables mainly from World Development Indicators (WDI).

### Endogeneity concerns and mitigation
- Reverse causality and omitted variables risks acknowledged.
- Mitigation strategies:
  - Cross-industry identification exploiting variation in ExtDep.
  - Regress average sectoral activity during crisis on pre-crisis bank positions.
  - Instrumental variable approaches (information sharing measures, strength of legal rights, legal origins, rule of law) and selection-on-observables sensitivity tests.

### 3. Main empirical findings

### Crisis period (2008–10) — core result (consistent with H1)
- Financially dependent industries in countries where banks had higher liquidity and capital ratios performed better during the crisis.
- Positive and statistically significant interaction between bank liquidity/capital (NSFR, Capital_Total, Capital_Tier1, PCA_ALL) and ExtDep for both:
  - Investment rate.
  - Establishment growth.
- Effects hold for bank-based financial systems and emerging market economies.

### Pre-crisis period (2000–07)
- Only a marginal link to bank capital is observed; liquidity results are not statistically significant.
- Effects are mostly evident for establishment growth rather than investment rate.
- Interaction term between Tier1 capital ratio (or PCA_ALL) and ExtDep is positive and statistically significant at the 10% level in some pre-crisis specifications.

### Quantitative illustrations (crisis period examples)
- Investment rate differentials (crisis; Column 9 (10) of Table 2A examples):
  - Moving from country at R 25th percentile to R 75th percentile, interaction coefficient alone implies industry at 90th percentile of ExtDep is 1.8% (1.9%) more than one at 10th percentile.
  - Including both interaction and R, differential ≈ 1.4% (1.3%).
  - Sample mean of investment rate: 9%; standard deviation: 12%.
- Establishment growth differentials (crisis; Column 9 (10) of Table 2B examples):
  - Interaction alone: industry at 90th percentile of ExtDep is about 6.6% (2.9%) more than one at 10th percentile when moving from R 25th to R 75th percentile.
  - Including coefficient on R: figures increase to 14% (12%).
  - Sample mean of establishment growth: 2%; standard deviation: 22%.

### Selected tabular magnitudes (preserved)
- Whole sample (Table 2A, investment rate, whole sample 2000-2010):
  - NSFR: R × ExtDep = 0.0274 (t = 1.41)
  - Capital_Total: R × ExtDep = 0.0019* (t = 1.94)
  - Capital_Tier1: R × ExtDep = 0.0018** (t = 2.13)
  - PCA_All: R × ExtDep = 0.0073** (t = 2.14)
- Whole sample (Table 2B, establishment growth, whole sample 2000-2010):
  - NSFR: R × ExtDep = 0.0942*** (t = 3.85)
  - Capital_Total: R × ExtDep = 0.0016 (t = 1.67)
  - Capital_Tier1: R × ExtDep = 0.0024*** (t = 3.08)
  - PCA_All: R × ExtDep = 0.0101*** (t = 4.84)

### 4. Robustness and identification

### Robustness checks summary
- Clustering:
  - Results robust to clustering at country level and two-way clustering (industry and country).
- Fixed effects:
  - Alternative specification with industry fixed effects and country-year fixed effects yields almost identical magnitudes.
- Instrumental variables (IV / 2SLS):
  - Instruments: information-sharing agency indicators, depth of credit information, strength of legal rights, legal origins, and rule of law.
  - Hansen tests generally fail to reject overidentification; KP-LM and first-stage F statistics indicate acceptable identification in many specifications.
  - IV estimates confirm baseline: crisis-period interactions remain positive and IV coefficients often larger than OLS.
- Sample and bank subsets:
  - Excluding too-big-to-fail banks (assets > USD 50 billion) yields similar results.
  - Focusing on banks more active in industrial loans (industrial-loans/total-loans > 50%) yields robust crisis-period findings and some pre-crisis marginal results.
  - Using only banks meeting Basel III minima: main capital findings robust; NSFR coefficients often not significant when using only banks meeting minima.
- Additional controls and channels:
  - Adding controls for bank stability (Zscore), bank competition (Boone), property rights, human development index (HDI), inflation, and interactions does not overturn crisis-period interaction results.
  - Selection-on-observables sensitivity analysis: computed ratios ∅3f/(∅3r−∅3f) suggest implausibility that results are fully due to omitted variables for crisis-period estimates.

### Addressing reverse causality
- Pre-crisis predictors:
  - Regressing crisis-period sectoral outcomes on pre-crisis bank positions (2000–07, 2005–07, 2006–07) shows consistent positive association for financially dependent industries located in countries with better pre-crisis liquidity and capital.
- Granger-style checks:
  - Across 48 regressions of country-level averages of bank liquidity/capital on average country-level sectoral activity, only two coefficients significant at 5% and three at 10% — little evidence sectoral activity Granger-causes bank liquidity/capital.
- Alternative IV and lag instruments:
  - Lagged R and R×ExtDep (one and two years) as instruments yield similar baseline results.

### 5. Policy implications, interpretation, and conclusions

### Main conclusions
- An increase in bank liquidity and capital is associated with stronger sectoral activity—higher investment rate and creation of new establishments—in financially dependent industries.
- This relationship is strong during the crisis period (2008–10) and only marginal in the pre-crisis period (2000–07).

### Policy interpretation
- Findings do not support the argument that Basel III-type reforms necessarily reduce economic activity by decreasing credit availability or raising borrowing costs.
- Evidence supports that bank buffers increase economy resilience to shocks, consistent with Basel III objectives.
- Macroeconomic costs of higher liquidity and capital requirements are likely negligible in terms of sectoral activity measured by investment and business expansion, lending support to tighter liquidity/capital standards.

### Caveats and suggested future research
- Results are based on relatively short windows for pre-crisis and crisis periods and have empirical limitations.
- Authors present results as notable correlations rather than definitive causal proof.
- Future research directions:
  - Establish a direct causal link from regulatory actions to economic outcomes.
  - Better understand transmission channels and why relationships differ between investment rate and establishment growth.

### 6. Key descriptive and sample statistics (preserved exactly)
- Bank-level observations: 1,857 banks.
- Country coverage for main regressions: 50 countries.
- Industry coverage: 28 industries.
- Time period: 2000–10.
- Investment rate (country-level average range): 1.4% (Colombia) to 45.1% (Georgia).
- Investment rate (industry-level average range): 6.0% (Wearing apparel, ISIC 322) to 15.3% (Glass products, ISIC 362).
- Investment rate (sample mean and standard deviation over 2000–10): mean = 9%, standard deviation = 12%.
- Establishment growth (country-level average range): -22.9% (Sri Lanka) to 14.6% (Vietnam).
- Establishment growth (industry-level average range): -3.9% (Misc. Petroleum and coal products, ISIC 354) to 5.7% (Fabricated metal products, ISIC 381).
- Establishment growth (sample mean and standard deviation over 2000–10): mean = 2%, standard deviation = 22%.
- Dependent variables winsorized at 5% tails.
- NSFR (country extremes): highest NSFR = 1.28 (Albania); lowest NSFR = 0.66 (Spain).
- Capital ratios (country extremes): Macedonia highest total = 29.4% and Tier 1 = 23.2%; Ireland lowest total = 10.9%; Morocco lowest Tier 1 = 8.8%.
- Share of banks meeting Basel III-style minima (Appendix Table A3):
  - NSFR ≥ 1: pre-crisis = 41%, crisis = 27% (precise figures in Appendix Table A3: 41.39% and 27.17%).
  - Capital_Total ≥ 10.5%: pre-crisis = 79%, crisis = 88% (precise: 79.47% and 87.99%).
  - Capital_Tier1 ≥ 8.5%: pre-crisis = 68%, crisis = 82% (precise: 68.28% and 81.88%).

*Source: wpiea2020103-print-pdf — 1. Introduction; 2. Background and Hypothesis Development; 3. Model Specification and Data; 4. Empirical Results; Appendix Table A2 — https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020103-print-pdf.pdf*

### 1.    Introduction

### 1.    Introduction

### Overview
- Financial crises impose large costs by requiring public funds to revive financial institutions and by destroying the finance-growth nexus.
- The global financial crisis in 2008–09 motivated a new wave of prudential regulations, including the Basel III Accord in 2010, aimed at improving financial stability and mitigating adverse impacts of financial shocks on the real economy.
- Basel III imposes more stringent micro-prudential standards for liquidity and capital positions and adds a macro-prudential overlay.

### Research question and empirical approach
- Main question: Do tighter bank regulations (higher liquidity/capital levels as in Basel III) affect sectoral economic activity, and does this effect differ between normal times and crisis periods?
- Identification strategy: Interaction approach inspired by Rajan and Zingales (1998) — interact a country’s bank-system soundness (liquidity and/or capital levels) with an industry’s external finance dependence to test whether sectors more reliant on external finance perform disproportionately better or worse in countries with higher bank liquidity/capital.
- Data and scope:
  - Bank-level dataset for 50 countries over the period 2000–10 to construct Basel III-style liquidity and capital indicators.
  - Merged with data on industry activity for 28 manufacturing sectors around the period of the global financial crisis.

### Key measures and outcome variables
- Two dimensions of sectoral activity:
  - Investment rate: gross fixed capital formation divided by output (following Calomiris et al. 2017).
  - Growth in the number of establishments: proxy for entrepreneurship and new business creation (following Claessens and Laeven 2003).
- Basel III-style bank indicators constructed: net stable funding, total regulatory capital, and Tier 1 ratios (computed using Basel III definitions rather than pre-existing liquid asset or quantity-based capital ratios).

### Theoretical expectations and channels
- Two competing effects of higher liquidity/capital:
  - Positive channel: Higher capital increases banks’ risk-bearing capacity, potentially creating more liquidity and more credit supply, especially business loans, supporting investment and firm creation.
  - Negative channel: Tighter requirements could lead banks to hoard liquidity or reduce lending to non-financial firms, lowering activity in the short run.
- The net effect may depend on macro-financial conditions:
  - During normal times, costs of tighter regulation may lower activity.
  - During crises, safer banks reduce market frictions, lower the cost of external finance, and can encourage investment and new firm formation.
- Sectors with higher external finance dependence are expected to be more sensitive to bank liquidity/capital levels; interaction term should capture this intensification.

### Main findings
- During the crisis period 2008–10:
  - External-finance-dependent industries in countries where banks had higher liquidity and capital ratios performed better.
  - Higher liquidity and capital levels (net stable funding, total regulatory capital, and Tier 1 ratios) are positively associated with higher investment rates and higher business creation in external-finance-dependent industries.
  - This positive association holds for bank-based financial systems and emerging market economies.
- During the pre-crisis period 2000–07:
  - Only a marginal link to bank capital is observed; liquidity results are not statistically significant.
  - Effects are mostly evident for establishment growth rather than investment rate.
- Quantitative illustration:
  - Moving from a country at the 75th percentile of the distribution of capital (liquidity) position to a country at the 25th percentile, the investment rate in more external-finance-dependent industries grew approximately 2% faster than their less dependent counterparts during the crisis, accounting for about 20% of the investment rate in this period.

### Robustness, identification, and interpretation
- Robustness checks:
  - Results robust to different econometric models and subsamples.
  - Controls for observable country/industry characteristics to mitigate omitted variables bias.
  - Selection on observables used to assess potential influence of unobserved heterogeneity.
- Addressing reverse causality and endogeneity:
  - Regressed average sectoral activity during the crisis on pre-crisis bank liquidity/capital values (pre-crisis bank positions cannot be affected by contemporaneous crisis activity).
  - Employed an instrumental variable approach to account for residual endogeneity concerns.
  - Tests indicate endogeneity is unlikely to overturn the established association.
- Caution: Authors interpret results as an informative correlation and view findings with care rather than claiming definitive causality.

### Policy implications and conclusions
- Findings do not support the argument that Basel III-type reforms necessarily reduce economic activity by decreasing credit availability or raising borrowing costs.
- Instead, evidence supports the argument that bank buffers increase economy resilience to shocks, consistent with Basel III objectives.
- The macroeconomic costs of higher liquidity and capital requirements are likely to be negligible in terms of sectoral activity measured by investment and business expansion, lending support to tighter liquidity/capital standards under Basel III.

### Contribution to the literature
- Adds to evidence that costs of Basel III are limited/transitory and that stable banks improve long-run growth.
- Advances the literature by using disaggregated sectoral data to investigate the channel through which higher bank capital/liquidity influences growth—by fostering entrepreneurship and firm creation in externally dependent sectors.
- Complements studies on bank performance and stability under Basel III and on the bank lending channel, emphasizing sustained lending by well-capitalized and well-funded banks during crises benefits investment and firm creation in financially dependent industries.

### Paper structure
- Section 2: Background on potential side effects and transmission channels of tougher liquidity and capital regulations.
- Section 3: Methodology, model specifications, and data description.
- Section 4: Results and related discussion.
- Section 5: Summary and conclusions.

*Source: wpiea2020103-print-pdf - 1.    Introduction — https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020103-print-pdf.pdf*

### 2. Background and Hypothesis Development

### 2. Background and Hypothesis Development

### Basel III objectives and trade-offs
- The Basel III Accord is a comprehensive set of reform measures for strengthening bank regulation, supervision, and risk management.
- Core elements: higher regulatory capital (both quantity and quality) and newly introduced liquidity requirements.
- Aim: promote a more flexible banking sector to absorb external shocks and decrease the risk of spillovers from the financial to the real sector.
- Ongoing debate: whether higher capital and liquidity requirements benefit the economy as a whole, because they may constrain banks’ intermediation role and contribution to economic activity.

### Four transmission channels through which higher capital requirements affect the economy
- (i) Reducing bank income
  - Substitution of debt with more expensive equity can reduce return on equity (ROE) because the decline in interest expenses will not be sufficient to maintain the higher return investors require on equity compared to debt.
  - To maintain returns, banks may increase their lending rates, which, given imperfect substitutability between bank credit and other financing, reduces aggregate credit supply and curbs economic activity (King 2010).
- (ii) Decreasing lending
  - Banks may shift asset portfolios and generate fewer loans to meet higher capital requirements (Fang et al. 2018).
  - Reduced profitability can force banks to increase lending rates, discouraging loan applications and lowering both supply of and demand for credit, curbing spending and investment (Miles et al. 2013; Furfine 2000; Roger and Vlček 2011; Boissary and Collard 2016; Fender and Lewrick 2016).
- (iii) Changing risk-taking behavior
  - Higher capital reduces incentives to take on more risk because potential losses to shareholders are larger in case of default.
  - Reduced bank risk-taking can shift demand for credit toward unregulated shadow banking, potentially increasing overall economy risk and exposure to financial crises. Offsetting forces may exist: shifting toward less risky assets reduces expected returns and charter value, possibly encouraging more risk-taking (Allen and Gale 2004).
- (iv) Reshaping competition
  - Higher capital may reduce the share of credit by large banks relative to smaller banks (large banks often hold smaller cushions above required ratios), reducing industry concentration and raising competition (D’Erasmo 2018).
  - Alternatively, tougher requirements can act as a barrier to entry over time, increasing incumbents’ market power, enabling higher lending rates and reducing economic performance (Hakenes and Schnabel 2011; Dagher et al. 2016).

- Net theoretical effect of stricter capital requirements on economic activity is ambiguous because these channels can work in opposing directions.

### Effects of tougher liquidity requirements
- Tougher liquidity requirements are costly because they discourage investment in risky assets.
- Higher liquidity requirements force banks to hold more low-yield liquid assets and long-term maturity funds, reducing bank revenue.
- To maintain targeted long-term ROE, banks would raise lending spreads (Dietrich et al. 2014; Kauko, 2017).
- Compliance shifts business strategies toward liability-driven asset management: find stable long-term funding first, then attempt to gain market share in lending (Allen et al. 2012).
- Potential negative effect: lending to productive projects may become inadequate if funding strategies shift.
- Potential positive effect: higher liquidity makes banks safer; perceived as partial deposit insurance by households, deposits may increase, enhancing core resources and promoting lending and economic activity (Agénor 2018).
- Liquidity and capital requirements tend to be complementary; e.g., a rise in liquid assets lowers risky assets per unit of equity, increasing equity’s disciplinary effect (Boissay and Collard 2016). However, they could interact unexpectedly, reducing overall benefits (Distinguin et al. 2013; Covas and Driscoll 2014; Cecchetti and Kashyap 2016; Goel et al. 2017).

### Empirical strategy: sectoral indicators and financial dependence
- Analysis captures potential effects of higher liquidity and capital standards on economic growth using two key sectoral activity indicators:
  - The investment rate.
  - Growth in the number of establishments.
- Sectors are distinguished according to how firms finance investments:
  - Internal funds (less dependent on external financing).
  - Reliance on external sources (more dependent on external financing).
- Motivation for indicators:
  - Investment rate: investment is a major channel through which bank regulations affect the economy. A lower investment rate may reflect firms’ inability to finance investment projects using external sources (bank loans). Tighter regulations affecting bank lending would more strongly affect sectors dependent on external financing.
  - Growth in number of establishments: one component of industry growth (the other is growth in average firm size). New establishment growth is less constrained by access to internal funds (existing establishments have access to internal funds), so growth in new establishments in externally dependent industries should be more sensitive to bank liquidity/capital regulations.

### Crisis versus normal times
- During financial crises, higher liquidity and capital requirements make banks more resilient, enabling them to sustain activities and support economic growth (Albertazzi and Marchetti 2010; Cornett et al. 2011; Puri et al. 2011; Beltratti and Stulz 2012; Kapan and Minoiu 2013).
- This support is expected to be stronger for more external-finance-dependent sectors (Popov and Udell 2012; Kroszner et al. 2007; Dell’Ariccia et al. 2008; Fernández et al. 2016; Moore and Mirzaei 2016).

### Main hypotheses
- H1: During a crisis, stricter liquidity regulation and higher capital requirements are positively associated with the investment rate and establishment growth, especially in external-finance-dependent sectors.
- H2: During normal times, the relationship between tougher regulations and activity indicators could be either positive or negative, or insignificant, and, if significant, more pronounced in external-finance-dependent sectors.

### Additional caveats and identification challenges
- Detecting a significant relationship between bank regulatory requirements and average sectoral activity may be difficult because the main transmission channel is bank lending.
- Better liquidity/capitalization may benefit the economy through channels (including consumption) not fully captured by sectoral patterns of business expansion; much literature uses aggregate measures such as GDP growth.
- Literature provides limited guidance on expected differences between the investment rate and establishment growth; the analysis lets the data speak to this point.

*Source: 2. Background and Hypothesis Development*

### 3.     Model Specification and Data

### 3.     Model Specification and Data

### Model specification: objective and empirical framework
- Aim: measure the effects of bank liquidity and capital requirements on sectoral activity (investment rate and creation of new businesses) in external-finance-dependent industries relative to others.
- Baseline empirical specification (Rajan and Zingales (1998) framework):
  - y_{i,c,t} = ϑ + ϕ1·Share_{i,c,t−1} + ϕ2·R_{c,t} + ϕ3·R_{c,t}×ExtDep_i + ϕ4·FinDev_{c,t} + ϕ5·FinDev_{c,t}×ExtDep_i + ε_{i,c,t}  (Eq.1)
  - y_{i,c,t}: sectoral activity measured as either gross fixed capital formation/output (investment rate) or growth in number of establishments (establishment growth).
  - Share: lagged sectoral share of value-added (controls for convergence; ϕ1 expected negative if convergence).
  - ExtDep: external finance dependence at the sectoral level (calculated using US firm-level data).
  - R_{c,t}: indicator of average bank liquidity or capital ratio in country c in year t.
  - Main parameter of interest: ϕ3 on R_{c,t}×ExtDep_i; a positive and significant ϕ3 implies financially dependent sectors perform relatively better in countries with higher bank liquidity/capital (consistent with hypothesis H1 for the crisis period). For the pre-crisis period, no particular sign of ϕ3 is predicted (H2).
  - ϕ2 captures the direct relationship between bank liquidity/capital and sectoral activity; moderate expectation of positive sign during the crisis and no particular sign pre-crisis.
- Controls for financial development and its interaction with ExtDep (FinDev and FinDev×ExtDep).
- Fixed effects: full set of sector, country, and year fixed effects (ϑ_i, ϑ_c, ϑ_t) as in Maskus et al. (2012).
- Estimation: OLS with standard errors clustered by industry; robustness checks include clustering at the country level and double-clustering at industry and country levels.

### Measurement of bank liquidity and capital (Basel III components)
- Focus on three central components of Basel III as measures of bank regulations:
  - NSFR (Net Stable Funding Ratio) as proxy for liquidity (chosen over LCR because more relevant for long-term performance and more straightforward to calculate with available historical data).
    - NSFR = Available stable funding (ASF) / Required stable funding (RSF) = (∑ z_i L_{ii}) / (∑ z_j A_{jj}), where L and A indicate liabilities and assets and z are weights between 0 and 1.
    - NSFR must be greater than 1 in Basel III.
    - Computation requires assumptions on ASF/RSF classifications and weights; Appendix Table A1 details components and factor weights.
  - Capital quantity: Capital_Total (total regulatory capital ratio, Tier 1 + Tier 2 to risk-weighted assets).
  - Capital quality: Capital_Tier1 (Tier 1 capital ratio to risk-weighted assets).
- Basel III calibration referenced:
  - Common Equity Tier 1 capital minimum increased from 4% to 4.5%; Tier 1 from 4% to 6%; overall regulatory capital unchanged at 8%.
  - Assumed minima in empirical work: at least 10.5% for total regulatory capital and 8.5% for Tier 1.
- Also use the first principal component of NSFR, Capital_Total, and Capital_Tier1 (PCA_ALL) in some regressions.

### Estimation issues: endogeneity and benchmarking
- Endogeneity concerns:
  - Reverse causality: banks that lend to faster-growing sectors may experience fewer loan losses and thus better capitalization, biasing inference.
  - Omitted variables: unobserved country-level factors (e.g., institutions, overall economic dynamism) may drive both bank health and firm/sector performance.
- Mitigation strategies:
  - Use of sectoral variation (cross-industry differences) to alleviate some endogeneity.
  - Instrumental variable (IV) approach and selection on observables approach are employed (details in Section 4.3).
- Benchmarking concern:
  - External-finance-dependence (ExtDep) is constructed using US firm-level data; mismeasurement bias may arise if industry features vary across countries.
  - IV strategy and sensitivity tests are used to assess robustness to benchmarking bias (see Section 4.2).

### Data: banks, industries, and countries
- Bank-level data:
  - Source: Bankscope.
  - Sample: data on 1,857 banks from 50 developed and emerging economies over the period 2000 to 2010.
  - Country-level bank measures are computed as unweighted averages of bank-level ratios across banks within a country.
  - Capital definitions:
    - Capital_Total: percentage of a bank's total capital (core Tier 1 + Tier 2) to risk-weighted assets.
    - Capital_Tier1: Tier 1 capital to risk-weighted assets.
  - NSFR computation follows Vazquez and Federico (2015) and Kapan and Minoiu (2013); Appendix Table A1 describes components and factor weights.
- Industry data:
  - Source: UNIDO Industrial Statistics Database (value added, output, number of establishments, gross fixed capital formation, employment).
  - Selected 73 industries of mixed 3&4-digit ISIC Rev. 3 codes, regrouped into 28 industries of ISIC Rev. 2 to match Rajan and Zingales (1998) external-finance-dependence classification.
  - UNIDO coverage: 135 countries, but sample restricted to countries with both industry activity variables and bank data for 2000–10; the United States is dropped (used for benchmarking), leaving 28 industries in 50 countries.
  - ExtDep for each industry is retrieved from Rajan and Zingales (1998) and based on US firm-level data.
- Country data:
  - Financial development (FinDev) and other country characteristics mainly from World Development Indicators (WDI).
  - Appendix Table A2 presents definitions and sources for all variables.

### Summary statistics and key descriptive facts
- Sample and coverage:
  - Bank-level observations: 1,857 banks.
  - Country coverage for main regressions: 50 countries.
  - Industry coverage: 28 industries (after regrouping).
  - Time period: 2000–10.
- Sectoral activity (investment rate and establishment growth):
  - Investment rate (country-level average range): 1.4% (Colombia) to 45.1% (Georgia).
  - Investment rate (industry-level average range): 6.0% (Wearing apparel, ISIC 322) to 15.3% (Glass products, ISIC 362).
  - Investment rate (sample mean and standard deviation over 2000–10): mean = 9%, standard deviation = 12%.
  - Establishment growth (country-level average range): -22.9% (Sri Lanka) to 14.6% (Vietnam).
  - Establishment growth (industry-level average range): -3.9% (Misc. Petroleum and coal products, ISIC 354) to 5.7% (Fabricated metal products, ISIC 381).
  - Establishment growth (sample mean and standard deviation over 2000–10): mean = 2%, standard deviation = 22%.
  - Dependent variables are winsorized at 5% in both tails to mitigate outliers.
- Bank liquidity and capital variation:
  - NSFR (country extremes): highest NSFR = 1.28 (Albania); lowest NSFR = 0.66 (Spain).
  - Capital ratios (country extremes): Macedonia highest total = 29.4% and Tier 1 = 23.2%; Ireland lowest total = 10.9%; Morocco lowest Tier 1 = 8.8%.
- Share of banks meeting Basel III-style minima (Appendix Table A3):
  - Liquidity: proportion of banks with NSFR ≥ 1 is about 41% in the pre-crisis period and 27% during the crisis.
  - Regulatory total capital ≥ 10.5%: 79% of banks in pre-crisis and 88% during the crisis.
  - Tier 1 capital ≥ 8.5%: 68% of banks in pre-crisis and 82% during the crisis.
  - Interpretation: most banks have capital ratios well above minima, consistent with precautionary buffers; empirical findings should be interpreted in terms of expectations of more stringent regulation rather than a direct test of Basel III minima.
- Descriptive patterns:
  - Figure 1 (described): bottom and top ten countries in sectoral activity over 2000–10 show emerging market economies among best-performing and most advanced countries among worst-performing.
  - Top-performing countries have, on average, more stable banking sectors than low-performing counterparts for both performance measures.
  - Figure 2 (described): linear fit of investment rate versus external finance dependence by NSFR group (>1 and <1) shows that during the crisis, sectors more dependent on external finance do proportionately better in countries with higher NSFR; difference evident only during the crisis period.

*Source: 3. Model Specification and Data, wpiea2020103-print-pdf*

### 4.     Empirical Results

### 4.     Empirical Results

### Baseline Results
- Sample periods analyzed: whole sample period (2000–10), pre-crisis period (2000–07), crisis period (2008–10).
- Dependent variables: investment rate (Table 2A) and establishment growth (Table 2B).
- Liquidity/capital measures used: NSFR, CapitalTotal, CapitalTier1, and PCA_ALL.
- Main crisis-period finding (consistent with H1):
  - During 2008–10, the interaction between bank liquidity/capital levels and external finance dependence is positive and statistically significant (columns 9–12), implying financially dependent sectors in countries with higher bank liquidity/capital had significantly higher investment rates and establishment growth.
- Pre-crisis period:
  - Little evidence that countries with high bank liquidity/capital performed better in 2000–07; only the interaction term between Tier1 capital ratio (or PCA_ALL) and external finance dependence is positive and statistically significant at the 10% level (columns 5–8).
- Direct links (main coefficients on R):
  - Investment rate: direct link between bank capital levels and investment rate is negative (columns 2–3), driven by the pre-crisis period (columns 6–7); no significant relationship during the crisis (columns 10–11). There is no significant direct link between bank liquidity levels and the investment rate in either subsample period.
  - Establishment growth: direct link between bank capital levels and establishment growth tends to be positive and significant in both pre-crisis and crisis periods (columns 6, 7, and 10). Direct link between bank liquidity levels and establishment growth is positive and significant only in the crisis period.
- Interpretation:
  - Higher bank liquidity/capital requirements could enhance prospects of young firms in externally finance-dependent industries by shielding them during crises—affecting the extensive margin (establishment growth) more than the intensive margin (investment by existing firms).
- Economic magnitudes (crisis period examples):
  - Investment rate differential (from Column 9 (10) of Table 2A):
    - Industry at 90th percentile of external finance dependence is 1.8% (1.9%) more than one at 10th percentile when moving from country at R 25th percentile to R 75th percentile using interaction coefficient alone.
    - When considering both interaction and R, the differential is approximately 1.4% (1.3%).
    - Sample mean of investment rate: 9%; standard deviation: 12%.
  - Establishment growth differential (from Column 9 (10) of Table 2B):
    - Industry at 90th percentile of external finance dependence is about 6.6% (2.9%) more than one at 10th percentile when moving from R 25th percentile to R 75th percentile using interaction coefficient alone.
    - When including coefficient on R, figures increase to 14% (12%).
    - Sample mean of establishment growth: 2%; standard deviation: 22%.
- Broader implication:
  - Strengthening liquidity and capital regulations could support sustainable growth and banking sector stability; well-capitalized banks can promote creation of new firms in normal and crisis times, and higher liquidity/capital can boost financially dependent sectors in bad times without hurting them in good times.

### Sensitivity Analysis
- Alternative clustering of standard errors:
  - Allowing correlations (i) across countries and (ii) across both sectors and countries (Petersen 2009 / Thompson 2011) — Tables 3A (country clustering) and 3B (sector-country clustering).
  - Results confirm baseline: investment rate and establishment growth in financially dependent sectors were significantly higher in countries with better bank liquidity/capital during the global financial crisis.
  - Coefficient on liquidity/capital itself is not significant at the 1% or 5% level when errors are clustered at country or sector-country level.
- Alternative fixed effects:
  - Use industry fixed effects (ϑi) and country-year fixed effects (ϑct) instead of industry, country, and year fixed effects (Table 3C). This requires dropping country-level variables (R and FinDev) and retaining interaction terms.
  - Coefficient magnitudes are almost identical to baseline; crisis-period findings remain intact; Tier 1–establishment growth link in pre-crisis becomes somewhat stronger.
- Alternative benchmark for external finance dependence:
  - IV estimator of Ciccone and Papaioannou (2010) to recover a country-independent proxy for external finance dependence; instrument Rct×ExtDep_i,US via a two-step strategy (predict industry slopes ∅̂i and construct ExtDep̂_i, then use Rct×ExtDep̂_i as instrument).
  - Regressions replicated in Table 3D: Kleibergen-Paap LM χ2 and first-stage F statistics generally do not indicate underidentification or weak instruments.
  - Main crisis-period findings robust; coefficients on interaction terms slightly larger than OLS (suggesting attenuation bias in OLS).
  - Marginal pre-crisis interaction coefficients (Tier 1) lose significance; total capital becomes marginally significant for establishment growth.
- Alternative channels controlled:
  - Add interactions with Tang (asset tangibility) and GO (industry sales growth in the US) to account for collateralizability and growth opportunities (Table 3E).
  - Interaction terms between all bank liquidity/capital ratios and external finance dependence remain positive and statistically significant during the crisis period; pre-crisis marginal association between Tier 1 and investment rate persists.

### Addressing Endogeneity Issues

- Omitted Variable Bias
  - Added six control variables and their relevant interactions (Table 4):
    1. Bank stability (Zscore) and Zscore×ExtDep.
    2. Bank competition (Boone indicator) and Boone×ExtDep.
    3. Property rights and property rights×sector-specific intangible intensity.
    4. Human development index and HDI×skill intensity.
    5. Interaction between bank liquidity/capital ratio and a proxy for liquidity.
    6. Inflation and inflation×ExtDep.
  - Results (Panel I pre-crisis; Panel II crisis):
    - Interaction term for all liquidity/capital ratios during the crisis remains positive and statistically significant.
    - Positive and statistically significant interaction for Tier 1 in the pre-crisis period.
  - Selection-on-observables sensitivity (Altonji et al. 2005; Nunn and Wantchekon 2011):
    - Compute ratio ∅3f/(∅3r−∅3f) comparing coefficients with restricted vs full controls.
    - All eight computed ratios for the crisis period, and four ratios for the pre-crisis period, are either greater than unity or negative (negative ratios imply controls strengthen results), suggesting implausibility that results are fully due to omitted variables.

- Reverse Causality
  - Concern: weak sectoral performance could raise bank NPLs, lowering bank capital/liquidity and causing observed correlations to reflect causality running from sectoral performance to bank health.
  - Strategy 1 — Cross-sectional pre-crisis predictors:
    - Regress average sectoral activity during crisis on pre-crisis averages of Share, R, and FinDev and their interactions with ExtDep, with sector and country fixed effects (Table 5A).
    - Findings: financially dependent industries performed better during crisis if located in countries where banks had better pre-crisis liquidity and capital positions. Robust when pre-crisis averages defined as 2000–07 (Panel I), 2005–07 (Panel II), and 2006–07 (Panel III); one exception: when averaging over 2000–07 the association between bank capital and establishment growth is positive but not significant.
  - Strategy 2 — Instrumental variables:
    - Instruments for bank liquidity/capital: information sharing measures and strength of legal rights in credit markets (World Bank “Doing Business” variables):
      - Existence of an information-sharing agency (public registry or private bureau) indicators and depth of credit information measure.
      - Strength of legal rights index for collateral and bankruptcy laws.
    - Rationale: information sharing and stronger creditor rights improve bank performance (profitability, lower risk) and thus liquidity/capital, but are unlikely to affect sectoral activity except via bank performance.
    - Also instrument FinDev and FinDev×ExtDep using legal origins and rule of law (literature-informed).
    - Exact variable definitions and sources provided in appendix (not reproduced here).

*Source: wpiea2020103-print-pdf — https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020103-print-pdf.pdf*

### Appendix Table A2.

### Appendix Table A2.

### Instrumental-variable results and identification tests
- Hansen test: fails to reject, at the 1% level, the validity of the overidentification exclusion restrictions (exception noted for the pre-crisis period value for the main interest variable).
- Instrumental-variable data availability: starting in year 2005.
- KP-LM 휒2 test and first-stage 퐹 statistics: show satisfactory values for underidentification and weak identification issues.
- IV estimator findings:
  - Confirms baseline results: financially dependent industries located in countries with banks with higher levels of liquidity and capital performed better during the global financial crisis.
  - Magnitude of IV coefficients is larger than those reported in Tables 2A and 2B for the OLS case.

### Additional robustness tests (section 4.4)
- Purpose: deepen understanding of how bank liquidity and capital are associated with sectoral activity.
- Excluding too-big-to-fail banks:
  - Definition: banks with total assets greater than USD 50 billion (following Berger and Bouwman (2013)).
  - Results: similar to Tables 2A and 2B; findings are not driven by a few big banks.
  - Alternative threshold robustness: results are robust if instead using a country-specific threshold excluding banks with market share greater than 10% (in terms of total assets).
- Focusing on banks more active in industrial loans:
  - Criterion: banks whose industrial-loans-to-total-loans ratio exceeds 50%.
  - Results: robust for the crisis period. In the pre-crisis period:
    - Positive association of Tier 1 capital with establishment growth.
    - Marginally significant positive association between bank liquidity and investment rate.
- Considering only banks that meet minimum liquidity and capital requirements:
  - Rationale: most banks meet minimums, so differences beyond minimum may not be meaningful.
  - Results (country-level aggregates using only banks meeting minimums): main findings regarding capital are robust.
  - NSFR result: coefficient on the NSFR is not significant — additional liquidity beyond a certain point is not associated with stronger investment or firm creation activity.
- System type and income-group split:
  - Method: split sample into market-based versus bank-based systems (depending on whether ratio of domestic credit to private sector to market capitalization is smaller or larger than the cross-country median) and advanced versus emerging market economies.
  - Expectation: ∅3 to be positive and significant for bank-based systems and for emerging market economies.
  - Results (Tables 7A and 7B): align with the conjecture.
    - The positive association between bank liquidity/capital and sectoral activity during the crisis is manifested only for countries in which banks play a more important role than the capital market.
  - Interpretation: bank-based systems (and emerging market economies) gain more benefits, in terms of sectoral activity, from tougher liquidity and capital requirements.

### Additional checks and supplementary evidence
- Granger-style checks (as applied by Berger and Sedunov (2017)):
  - Across 48 regressions of country-level averages of bank liquidity/capital on average country-level sectoral activity, only two coefficients are statistically significant at the 5% level, and three more at the 10% level — suggesting no strong evidence that sectoral activity Granger-causes bank liquidity/capital positions.
- Alternative IV strategy:
  - Use lags (one and two) of 푅 and 푅×퐸푥푡퐷푒푝 as instruments; results similar to baseline (not reported for brevity).
- OLS lag check:
  - Applying OLS with a one-year lag of 푅 yields quantitatively similar results to baseline.

### Main conclusions (section 5)
- Core empirical finding:
  - An increase in bank liquidity and capital is associated with stronger sectoral activity in the form of a higher investment rate and the creation of new establishments in financially dependent industries.
  - This relationship is strong during the crisis period, but only marginal in the pre-crisis period.
- Policy interpretation and caveats:
  - Findings may justify the argument that banks can sustain lending and support economic activity during a crisis if they maintain appropriate liquidity and capital levels, without significant negative consequences for lending and economic activity in normal times.
  - Caution: conclusions are based on relatively short windows for pre-crisis and crisis periods and other empirical limitations; findings are presented as a notable correlation rather than definitive causal proof.
- Suggested directions for future research:
  - Establish a causal link directly from regulatory actions to economic outcomes.
  - Better understand transmission channels and why relationships differ between investment rate and establishment growth.

*Source: Appendix Table A2, wpiea2020103-print-pdf - Appendix Table A2.*

### References

### References and Empirical Appendix — Key Results and Data Definitions

### Major empirical findings
- R × ExtDep (interaction of bank liquidity/capital indicator R with industry external financial dependence) is positive and often statistically significant for sectoral activity measures:
  - Table 2A (Investment rate, whole sample 2000-2010)
    - NSFR: R × ExtDep = 0.0274 (t = 1.41)
    - Capital_Total: R × ExtDep = 0.0019* (t = 1.94)
    - Capital_Tier1: R × ExtDep = 0.0018** (t = 2.13)
    - PCA_All: R × ExtDep = 0.0073** (t = 2.14)
    - Differential in sectoral activity (%) without direct effect of R: 1.09, 0.95, 1.33, 1.00 (columns [1]–[4])
    - Differential in sectoral activity (%) with direct effect of R: 0.46, 0.55, 0.65, 0.31 (columns [1]–[4])
  - Table 2B (Establishment growth, whole sample 2000-2010)
    - NSFR: R × ExtDep = 0.0942*** (t = 3.85)
    - Capital_Total: R × ExtDep = 0.0016 (t = 1.67)
    - Capital_Tier1: R × ExtDep = 0.0024*** (t = 3.08)
    - PCA_All: R × ExtDep = 0.0101*** (t = 4.84)
    - Differential in sectoral activity (%) without direct effect of R: 1.63, 1.27, 1.85, 0.87 (columns [1]–[4])
    - Differential in sectoral activity (%) with direct effect of R: 1.82, 2.50, 3.55, 2.92 (columns [1]–[4])
- Effects are generally larger and more statistically robust during the crisis period (2008–2010) than in the pre-crisis period (2000–2007):
  - Table 2A (Investment rate): columns labeled “Crisis (2008–2010)” show differentials (without/with direct effect) up to 1.92 and 1.34 (see Table 2A bottom rows and columns [9]–[12]).
  - Table 2B (Establishment growth): crisis-period differentials (with direct effect) up to 11.31 (columns [9]–[12], crisis rows).
- Financial development (FinDev) has a small positive main coefficient in many specifications for investment rate:
  - Table 2A: FinDev = 0.0001*** (t = 3.91) and similar magnitudes across columns and samples.
- FinDev × ExtDep is often negative and statistically significant for establishment growth in several robustness specifications (e.g., Table 3B, crisis columns show FinDev × ExtDep = -0.0001*** to -0.0003*** with strong t-statistics).

### Robustness and alternative specifications
- Robust to clustering choices:
  - Table 3A (cluster at country level) and Table 3B (two-way clustering at industry and country) show R × ExtDep coefficients remain positive and in many cases significant (e.g., Table 3B, Crisis investment rate: R × ExtDep = 0.1118* (t = 1.95) in column [1]; establishment growth R × ExtDep = 0.4102*** (t = 3.62) in columns [5]–[8]).
- Robust to alternative fixed effects:
  - Table 3C (sector and country*year fixed effects) reports R × ExtDep = 0.1119** (t = 2.46) for investment rate in one specification and R × ExtDep = 0.3994*** (t = 3.53) for establishment growth in another; differential in sectoral activity increases (e.g., 2.06, 2.42, 6.47 in specified cells).
- Instrumental variables / 2SLS approaches:
  - Table 3D and Table 5B use 2SLS/IV (instruments: information-sharing, legal rights, legal origins, rule of law and interactions). Selected IV results:
    - Table 3D (2SLS, pre-crisis / crisis panels): R × ExtDep = 0.4584** (t = 2.24) in one specification; first-stage F tests for R × ExtDep range from 1.64 to 7.57 (see table rows “1st Stage F test R × ExtDep”).
    - Table 5B (IV, whole sample 2005–2010): NSFR specification R × ExtDep = 0.0257 (t = 0.22) in one column, and for alternative splits and periods R × ExtDep up to 0.2212*** (t = 2.78) or 0.5315*** (t = 3.50) in other columns; high 1st-stage F statistics reported (e.g., R: 551***, R × ExtDep: 2322*** in some panels).
    - Hansen J p-values and KP-LM χ2 p-values reported in Table 5B confirm instrument tests in many columns (e.g., Hansen J p-value = 0.11, KP-LM χ2 p-value = 0.000 for some estimations).
- Robust to alternative samples and bank subsets:
  - Table 6A–C present robustness excluding “too big to fail” banks, excluding banks less important for industrial loans, or only banks meeting Basel III thresholds. Key selected figures:
    - Table 6A (excluding 130 “too big to fail” banks): establishment growth R × ExtDep = 0.3693*** (t = 3.90) in one crisis specification.
    - Table 6B (excluding banks not reporting industrial loans): investment rate R × ExtDep = 0.0263* (t = 2.00) in pre-crisis sample; establishment growth R × ExtDep = 0.1482*** (t = 4.15) in crisis sample.
    - Table 6C (banks meeting Basel III thresholds NSFR>=1, Capital_Total>=10.5%, Capital_Tier1>=8.5%): mixed R × ExtDep signs and significance across samples; in establishment growth crisis panels, R × ExtDep = 0.0925 (not always significant) and other columns show 0.0062** or 0.0082**.
- Robust to controlling for additional channels and observables:
  - Table 3E, Table 4, Table 5A, Table 7 and Table 4 (reverse causality tests) include interactions with tangible intensity (Tang), growth opportunities (GO), bank stability, competition, property rights, HDI, inflation and provide further checks. Examples:
    - Table 3E (other channels): R × ExtDep = 0.0213 (t = 1.11) in one investment specification; R × ExtDep = 0.4333*** (t = 3.20) in an establishment-growth PCA_All crisis specification.
    - Table 4 (reverse causality): lagged R coefficients are often significant (e.g., R(t-n) = 0.4386*** (t = 2.93) in column [1]), while sectoral activity lags generally do not Granger-cause R in a consistent manner.

### Magnitudes and selected summary statistics (preserved exactly as reported)
- Panel D summary statistics (Table 1):
  - Investment rate: N = 11058, Mean = 0.0900, S.D. = 0.1200, Min. = 0.0100, 0.25 = 0.030, Mdn. = 0.050, 0.75 = 0.090, Max. = 0.62
  - Establishment growth: N = 11907, Mean = 0.0200, S.D. = 0.220, Min. = -1.000, 0.25 = -0.05, Mdn. = 0.000, 0.75 = 0.05, Max. = 1.99
  - NSFR: N = 5100, Mean = 0.9500, S.D. = 0.1800, Min. = 0.1700, 0.25 = 0.850, Mdn. = 0.93, 0.75 = 1.04, Max. = 1.91
  - Capital_Total (%): N = 4721, Mean = 7.05, S.D. = 4.98, Min. = 9.06, 0.25 = 13.22, Mdn. = 16.29, 0.75 = 19.49, Max. = 39.60 (note: table cells preserve ordering as in source)
  - Capital_Tier1 (%): N = 4431, Mean = 3.61, S.D. = 4.58, Min. = 6.40, 0.25 = 10.23, Mdn. = 12.56, 0.75 = 16.06, Max. = 31.08
  - FinDev (%): N = 5501, Mean = 34.28, S.D. = 91.64, Min. = 4.67, 0.25 = 5.94, Mdn. = 27.94, 0.75 = 127.70, Max. = 508.42
- Table A3 (stable vs risky banks, averaged)
  - NSFR >= 1: Pre-crisis (2000–2007) = 41.39%, Crisis (2008–2010) = 27.17%
  - Capital_Total >= 10.5%: Pre-crisis = 79.47%, Crisis = 87.99%
  - Capital_Tier1 >= 8.5%: Pre-crisis = 68.28%, Crisis = 81.88%
  - Complementary shares: NSFR < 1 = 58.61% pre-crisis, 72.83% crisis; Capital_Total < 10.5% = 20.53% pre-crisis, 12.01% crisis; Capital_Tier1 < 8.5% = 31.72% pre-crisis, 18.12% crisis.

### Variable definitions and key data construction notes
- Sectoral activity outcomes:
  - Investment rate = ratio of gross fixed capital formation to output (UNIDO database, own calculation; nominal output deflated using producer price index of finished goods).
  - Establishment growth = simple growth rate of number of establishments (UNIDO database, own calculation).
- Industry characteristics:
  - Share = sector value added as share of total value added (UNIDO, own calculation).
  - ExtDep = external financial dependence from Rajan and Zingales (1998), industry-level U.S. median of (capex − cash flow)/capex by 3-digit ISIC.
  - Tang = tangibility proxy (industry median net PPE / total assets, Kroszner et al. 2007).
  - GO = growth opportunities proxy (industry median growth in real sales for U.S. firms, Ciccone & Papaioannou 2006).
- Bank liquidity and capital measures R:
  - NSFR = net stable funding ratio following Basel III weights (computed in line with Basel III proposal; Table A1 shows assets and liabilities items and weights; source: Bankscope and own calculation).
  - Capital_Total = regulatory total capital ratio (Bankscope).
  - Capital_Tier1 = regulatory Tier 1 capital ratio (Bankscope).
  - PCA_All = first principal component combining NSFR, Capital_Total, Capital_Tier1.
- Financial development (FinDev) = sum of domestic credit to private sector and market capitalization as % of GDP (World Bank-WDI, own calculation).
- Controls and instruments:
  - Bank Stability (World Bank Global Financial Development Database), Bank Competition (Lerner index from World Bank), Property Right (Heritage Foundation Database), Credit Bureau/public/private, Strength of legal rights, Legal Origin, Rule of Law (World Bank/WGI), among others listed in Table A2.

### Empirical formats, samples, and inference
- Sample: 28 industries (3-digit ISIC Rev.2) across 50 countries; period typically 2000–2010, with some analyses using 2005–2010 for IVs.
- Fixed effects: baseline specifications include sector, country, and year fixed effects; alternative specs use sector and country*year fixed effects (see Table 3C).
- Standard errors: robust standard errors clustered at industry level (baseline); sensitivity reported for clustering at country level and two-way clustering (industry and country).
- Significance notation: ***, **, * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

*Italicized source: References and tables from the PDF content unit wpiea2020103-print-pdf (References, Tables 1–A4, Figures 1–2, Appendix tables A1–A3).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020103-print-pdf.pdf_