## _wp15145

## Source details

**Canonical URL:** [_wp15145](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15145.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15145.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15145.pdf.json)

---

### INTRODUCTION — background, measures, and data
- Global financial crisis of 2007-08 led to a chronic shortfall of bank capital and a sustained decline in private investment.  
- Private investment in advanced economies declined by an average of 25 percent (see IMF World Economic Outlook, April 2015).  
- Two measures of bank capital adequacy:
  - Tier 1 ratio: ratio of Tier 1 capital to risk-weighted assets (traditional regulatory capital ratio).  
  - Leverage ratio: ratio of Tier 1 capital to total assets (the Basel III leverage ratio).  
  - Basel III minimum leverage ratio: 3 percent globally.  
  - U.S. authorities announced raising the minimum leverage ratio to 6 percent for systemically important U.S. banks (July 2013).  
- Key sectoral mechanism studied:
  - Hypothesis: Tier 1 ratio biases banks’ asset holdings toward government bonds and away from corporate and household credit (higher risk weights on lending), while a higher simple leverage ratio may support broader lending.
- Data scope:
  - Sample: 11,106 non-financial firms in sixteen advanced economies (post-crisis period).  
  - Banking coverage: 1,032 listed banks across those 16 countries (Bankscope).  
  - Firm-year observations: 48,835 firm-year pairs.  
  - Median investment-to-capital ratio: 20 percent.  
  - Average banking-system leverage ratio: 5 percent across country-years.  
  - Average Tier 1 capital ratio: 11 percent during this period.  
  - Example: EU commercial banks increased government debt holdings from 1.75 trillion euros in 2007 to 2.76 trillion euros in 2013, accounting for 13.56% of GDP in 2007 and 20.45% of GDP in 2013 respectively.

### EMPIRICAL STRATEGY AND CONTROLS
- Financial dependence measure:
  - Sector-level “Financial Dependence” (Rajan-Zingales style index) constructed from U.S. firms (1990-2006) and expanded to 90 SIC 3-digit sectors.
- Key explanatory variables:
  - Lagged country-level leverage ratio (asset-weighted), lagged Tier 1 ratio (asset-weighted), lagged nonperforming loans ratio, and bank-vs-market-based financial system index (bank credit to private sector over stock market capitalization).
- Controls and identification:
  - Firm fixed effects; country-year-sector fixed effects; firm-level sales-to-capital ratio; Tobin’s Q defined as (Market capitalization + Total Assets – Total Equity)/Total Assets.
  - Leverage ratio and Tier 1 ratio series are demeaned by subtracting a global median for each year; one-year lags used.

### MAIN RESULTS (summary)
- Directional findings:
  - Firms with greater external financial needs invest relatively more in countries that have higher bank leverage ratios.
  - Leverage ratio interaction with financial dependence is positive and statistically significant at the 5 percent level in baseline regressions.
  - Tier 1 ratio interaction with financial dependence is positive but not statistically significant in baseline cross-country setting.
  - Nonperforming loans ratio interacted with financial dependence is not statistically important in baseline specifications.
  - The relative importance (size) of the banking sector (bank-vs-market based index) interacted with financial dependence is not significant.
  - Interaction of lag real GDP growth with financial dependence is significant at the 5 percent level; leverage*financial-dependence remains significant when controlling for this.
- Robustness:
  - Result robust to inclusion of Tier 1 ratio and nonperforming loans alongside leverage ratio.
  - Robust to sample changes, sector restrictions, inclusion of additional controls, and alternative constructions of the bank-based financial system index.

### QUANTITATIVE MAGNITUDES (examples cited)
- Sector comparison: the “Engines and Turbines” sector in the U.S. (at the top decile of bank leverage ratio) will invest 9.7% more than the same sector in France (at the bottom decile of bank leverage ratio).
- Subsample of firms with sound fundamentals (see below): coefficients on interactions between bank capital measures and financial dependence lie between 2.226 and 2.735 (subsample) versus 0.830 to 1.570 (full sample).
- Macro illustrative magnitudes (from conclusions):
  - Going from the average level of capital in U.S. banks in 2007 to the level that prevailed in 2013, holding all else constant, had the effect of adding around 0.56 percent of GDP to corporate investment.
  - If the U.S. were today to have the average level of Tier 1 capital-to-total assets prevailing among European banks, it would result in 0.73 percent of GDP lower investment, purely from the effect that lower bank capital has on the availability of credit to firms reliant on external financing.

### 5.1 Firms with “good fundamentals” — subsample evidence
- Subsample construction:
  - Firms in the lowest one-third of the sample for expected default risk based on the O-score (one-year-ahead probability of default).
  - Observations reduced from about 46,000 to 18,000.
  - Median O-score used to determine cutoff for the lowest one-third to ensure firm type stability across years.
- Key findings for this subsample:
  - The significance of the bank’s leverage ratio for financially dependent firms persists.
  - Coefficient magnitudes on capital*financial-dependence interactions increase: between 2.226 and 2.735 (subsample) versus 0.830 to 1.570 (full sample).
  - In Column 3: Tier 1 ratio × financial dependence becomes weakly significant at the 10 percent level but much smaller in magnitude than leverage interaction.
  - In Column 5 (both leverage and Tier 1 included): leverage × financial dependence remains significant at the 1 percent level; Tier 1 × financial dependence loses significance.
  - Interpretation: bank capital constraints on credit supply are more binding for financially dependent firms with sound fundamentals.

- O-score definition (formula and components preserved exactly):
  - O-Score =
    - Total Liability/Total Asset * 1.32
    - + Working Capital/Total Asset * 0.41
    - + Current Liabilities/Total Asset * 6.03
    - - Size * 1.43
    - + Current Asset/Total Asset * 0.08
    - - Net Income/Total Asset * 2.37
    - + FFO/Total Asset * 1.83
    - + F * 0.28
    - - G * 1.72
    - + H * 0.52
  - Definitions:
    - Size is the log of total asset divided by the GDP deflator.
    - FFO means pre-tax income plus depreciation and amortization.
    - F is a dummy equal to one if cumulative net income over the previous two years is negative.
    - G is a dummy equal to one if owners’ equity is negative.
    - H is the change in net income.

### 5.2 Sector-level regression — aggregation robustness
- Motivation: address potential bias from large share of U.S. firms by constructing sector-country-year observations at 3-digit SIC level.
- Sample and construction:
  - Every variable constructed at the sector-country-year level.
  - New independent variable: average investment-to-capital ratio of firms in the top tercile of fundamentals (lowest expected probability of default).
  - Share of U.S. observations in this sector-level sample drops to 25%.
- Findings:
  - Sector-level results broadly unchanged from firm-level results.
  - Including sectoral average sales-to-capital ratio: sectors with higher average sales-to-capital ratio invest more, and leverage × financial-dependence remains significant.
  - Tier 1 ratio, nonperforming loans ratio, and importance of bank-based financial system are statistically insignificant in sector regressions.
  - Dropping U.S. sectors (reduces sample size by 25%): equity/assets × financial dependence remains significant at the 5% level with coefficient 1.24 (slightly smaller than whole-sample 1.36).
  - Conclusion: key findings are not driven only by U.S. samples.

### POLICY-RELEVANT IMPLICATIONS AND CONCLUSIONS
- Core empirical conclusion:
  - Firms with larger external financing needs invest relatively more in countries that are able to maintain higher bank leverage ratios; effect stronger for firms with sound fundamentals.
- Policy implications:
  - Countries proactive in addressing banking sector capital shortfalls and providing regulatory and other incentives for banks to raise capital (in the sense of a higher leverage ratio) see dividends in the form of higher corporate investment ratios.
  - Firms with sound fundamentals and solid investment projects are less likely to be credit-constrained when bank balance sheets are stronger.

*Source: _wp15145*

### INTRODUCTION

### INTRODUCTION

### Background and motivation
- Global financial crisis of 2007-08 led to a chronic shortfall of bank capital and a sustained decline in private investment.  
- Private investment in advanced economies declined by an average of 25 percent (see IMF World Economic Outlook, April 2015).  
- During the crisis banks faced losses on loan portfolios and increases in regulatory capital requirements, constraining their ability to raise capital.  
- Balance sheet weaknesses and scarcity of bank-intermediated credit contributed to lower corporate investment.  
- Extensive literature documents how adverse shocks to bank capital can affect lending supply and corporate investment (Bernanke and Lown (1991); Peek and Rosengren (1997, 2000); Calomiris and Mason (2003)).  
- April 2015 IMF World Economic Outlook finds financially constrained firms—those more reliant on financing from banks and capital markets—were more likely to have lower investment after adjusting for other factors (see also Claessens, Tong and Wei, 2012).

### Two measures of bank capital adequacy
- Tier 1 ratio: ratio of Tier 1 capital to risk-weighted assets (traditional regulatory capital ratio).  
- Leverage ratio: ratio of Tier 1 capital to total assets (the Basel III leverage ratio).  
- Basel III minimum leverage ratio: 3 percent globally.  
- U.S. authorities announced raising the minimum leverage ratio to 6 percent for systemically important U.S. banks (July 2013).  
- Short-run bank responses to capital requirements often involve shifting asset composition toward lower-risk assets (largely sovereign debt). Over a longer horizon banks can build capital via capital markets to expand corporate lending.

### Key sectoral mechanism studied
- Hypothesis: Tier 1 ratio biases banks’ asset holdings toward government bonds and away from corporate and household credit (higher risk weights on lending), while a higher simple leverage ratio may support broader lending.  
- Empirical strategy focuses on heterogeneity across firms in external financial dependence and examines how leverage ratio versus Tier 1 ratio affect investment by financially dependent firms.

### Data scope
- Sample: 11,106 non-financial firms in sixteen advanced economies (post-crisis period).  
- Banking coverage: 1,032 listed banks across those 16 countries (Bankscope).  
- Firm-year observations: 48,835 firm-year pairs.  
- Median investment-to-capital ratio: 20 percent.  
- Average banking-system leverage ratio: 5 percent across country-years.  
- Average Tier 1 capital ratio: 11 percent during this period.  
- Example of sovereign-holding shift: commercial banks in the European Union increased government debt holdings from 1.75 trillion euros in 2007 to 2.76 trillion euros in 2013, which accounts for 13.56% of GDP in 2007 and 20.45% of GDP in 2013 respectively.

### Main empirical approach (overview)
- Use sector-level “Financial Dependence” (Rajan-Zingales style index) to classify firms’ intrinsic dependence on external finance for investment; RZ index constructed from U.S. firms (1990-2006) and expanded to 90 SIC 3-digit sectors.  
- Key explanatory variables: lagged country-level leverage ratio (asset-weighted), lagged Tier 1 ratio (asset-weighted), lagged nonperforming loans ratio, and bank-vs-market-based financial system index (bank credit to private sector over stock market capitalization).  
- Controls: firm fixed effects; country-year-sector fixed effects; firm-level sales-to-capital ratio; Tobin’s Q defined as (Market capitalization + Total Assets – Total Equity)/Total Assets.  
- Detrending: leverage ratio and Tier 1 ratio series are demeaned by subtracting a global median for each year to control for global increases post-2008; one-year lags used to reduce endogeneity.

### Paper organization
- Section 2: literature review.  
- Section 3: methodology and data.  
- Section 4: baseline results.  
- Section 5: robustness checks.  
- Section 6: conclusion.

---

### Main results (summary bullets)
- Firms with greater external financial needs invest relatively more in countries that have higher bank leverage ratios.  
  - This result is robust to inclusion of other explanatory variables (system average nonperforming loans ratio; share of intermediation by banks vs capital markets).  
- Magnitude example: the “Engines and Turbines” sector in the U.S. (at the top decile of bank leverage ratio) will invest 9.7% more than the same sector in France (at the bottom decile of bank leverage ratio).  
- Effect is stronger when restricting the sample to the top one-third of firms in terms of underlying fundamentals (as measured by their one year ahead probability of default).  
- Empirical findings are robust to changing the sample of countries and restricting to a subset of sectors.  
- Leverage ratio interaction with financial dependence is positive and statistically significant at the 5 percent level in baseline regressions.  
- Tier 1 ratio interaction with financial dependence is positive but not statistically significant—suggesting Tier 1 ratio does not appear to have a significant effect on corporate investment in this cross-country setting, possibly because higher Tier 1 ratios capture a high share of low risk-weight assets.  
- Nonperforming loans ratio interacted with financial dependence is not found to be statistically important in baseline specifications.  
- Including interactions for Tier 1 ratio and nonperforming loans alongside leverage ratio does not alter the positive and significant leverage*financial-dependence result.  
- The relative importance (size) of the banking sector (bank-vs-market based index) interacted with financial dependence is not significant—implying quality of banking sector matters more than relative size.  
- Interaction of lag real GDP growth with financial dependence is significant at the 5 percent level (economic recovery boosts investment), and leverage*financial-dependence remains significant when controlling for this.

### Evidence from related literature (select findings cited)
- Aiyar, Calomiris and Wieladek (2014): increase in bank capital requirement ratio of 100 basis points reduces lending growth by 5.7 percentage points (U.K., 1998–2007).  
- Brun, Fraisse and Thesmar (2013): French transition from Basel I to Basel II reduced average capital requirement by about 2 percentage point (from 8% to 6%), leading to an increase in average loan size by approximately 10% (loan-level data 2006–2011).  
- Chiuri, Ferri and Majnoni (2002): enforcement of CARs in 15 emerging economies significantly reduced credit supply, especially for less-well capitalized banks.  
- Kashyap, Stein and Hanson (2010): argue substantially higher capital requirements will have a limited impact on loan costs; estimate that for a 2.5 basis-point increase in loan rates, a ten percentage point increase in the capital ratio would be required.  
- Goodhart, Sunirand, and Tsomocos (2006): higher bank capital may reduce bank-run risk, improve credit ratings, lower funding costs and lending rates, and enable banks to increase holdings of risky assets like loans.  
- Carlson, Shan and Warusawitharana (2012); Cornetta et al. (2011): banks with higher pre-crisis capital or more equity financing showed relatively higher lending during the 2007–09 crisis.  
- Kapan and Minoiu (2015): in U.S. bank loan deals (2006–2010), loan deals unrelated to Tier 1 ratio but positively related to bank equity to tangible assets ratio—consistent with this paper’s cross-country results.

### Key sample and descriptive statistics
- Countries: 16 advanced countries in North America and Europe experiencing financial crisis in 2008 or 2009 (Laeven and Valencia (2012) database).  
- Firm-level data source: Worldscope; total non-financial firms in sample: 11,106.  
- Bank-level data source: Bankscope; sample includes 1,032 listed banks.  
- Firm-year observations: 48,835.  
- Median investment-to-capital ratio: 20 percent.  
- Significant heterogeneity: investment ratio at the 75th percentile of firms (by financial dependence) is four times larger than at the 25th percentile.  
- Average leverage ratio of bank system: 5 percent across country-years.  
- Average Tier 1 capital ratio: 11 percent during the sample period.

### Empirical specifications and robustness checks (high-level)
- Baseline regressions interact country-level leverage ratio (lagged, detrended) with sector-level financial dependence index; firm fixed effects and country-year-sector fixed effects included.  
- Alternative specifications: add firm-level sales-to-capital and Tobin’s Q; replace leverage ratio with Tier 1 ratio; interact nonperforming loans ratio with financial dependence; include both leverage and Tier 1 interactions; add bank-vs-market-based financial system interaction; control for lag real GDP growth.  
- Robustness: results robust to sample changes, sector restrictions, inclusion of additional controls, and alternative constructions of the bank-based financial system index (bank credit to private sector over stock and bond market capitalization).

*Source: _wp15145 - INTRODUCTION*

### 5.1 Firms with “good fundamentals”

### 5.1 Firms with “good fundamentals”

### Subsample selection and methodology
- Subsample consists of firms in the lowest one-third of the sample for expected default risk.
- The one-year-ahead probability of default is measured using the O-score.
- The number of observations in this subsample is reduced from about 46,000 to 18,000.
- Regressions identical to the main analysis are run on this subsample to test whether weak firm fundamentals, rather than bank-intermediated credit supply, drive investment outcomes.

### Key empirical findings
- The significance of the bank’s leverage ratio for financially dependent firms persists in the subsample of firms with sound fundamentals.
- The size of the coefficients on the interaction between bank capital measures and financial dependence is larger in this subsample:
  - Coefficients are between 2.226 and 2.735 in the subsample.
  - This compares to the range from 0.830 to 1.570 reported in Table 3 for the full sample.
- In Column 3:
  - The interaction between Tier 1 ratio and financial dependence becomes weakly significant at the 10 percent level.
  - The magnitude of this Tier 1 interaction is much smaller than that for the interaction between the leverage ratio and financial dependence.
- In Column 5 (when both leverage ratio and Tier 1 ratio are included):
  - The interaction between the leverage ratio and financial dependence remains significant at the 1 percent level.
  - The interaction between the Tier 1 ratio and financial dependence loses its significance.
- Interpretation: the level of bank capital appears even more relevant for financially dependent firms with sound fundamentals, suggesting that constraints on the supply of bank credit are likely to be more binding for such firms’ investment outcomes.

### O-score definition and operationalization
- The O-score combines nine accounting ratios into a single statistic. The formula as presented:
  - O-Score =
    - Total Liability/Total Asset * 1.32
    - + Working Capital/Total Asset * 0.41
    - + Current Liabilities/Total Asset * 6.03
    - - Size * 1.43
    - + Current Asset/Total Asset * 0.08
    - - Net Income/Total Asset * 2.37
    - + FFO/Total Asset * 1.83
    - + F * 0.28
    - - G * 1.72
    - + H * 0.52
  - Where:
    - Size is the log of total asset divided by the GDP deflator.
    - FFO means pre-tax income plus depreciation and amortization.
    - F is a dummy equal to one if cumulative net income over the previous two years is negative.
    - G is a dummy equal to one if owners’ equity is negative.
    - H is the change in net income.
- To ensure firms do not change their types across sample years, the median O-score is denoted as firms’ O-score and used to determine the cutoff for the lowest one-third.

*Source: 5.1 Firms with “good fundamentals”*

### 5.2 Sector level regression

### 5.2 Sector level regression

### Motivation and sample construction
- Concern: the share of U.S. firms in the firm-level dataset is relatively large and U.S.-specific factors could bias results.
- Mitigation: regressions re-run at the 3-digit SIC sector-country-year level; every variable constructed at the sector-country-year level.
- New independent variable: the average investment to capital ratio of the firms in the top tercile of fundamentals (lowest expected probability of default).
- In this sector-level sample, the share of U.S. observations drops to 25%.
- Illustration: at the firm level, there are 61 U.S. firms in the “Drugs” sector versus 42 firms in the other 11 European countries; compressed at the sector level this yields 1 U.S. observation and 11 observations for the other European countries.

### Main sector-level regression findings
- Table 5: sector-level results broadly unchanged from firm-level results (Column 1).
- Including sectoral average sales-to-capital ratio as a control (Column 2):
  - Sectors with higher average sales-to-capital ratio are more likely to invest more.
  - Even after including this control, the interaction of bank leverage ratio with the financial dependence index remains significant.
- Sensitivity analyses:
  - Including Tier 1 ratio, nonperforming loans ratio, and the importance of bank-based financial system: those variables are statistically insignificant.
- Robustness check: dropping U.S. sectors (Column 8 of Table 5)
  - Reduces sample size by 25%.
  - Interaction of equity/assets and financial dependence remains significant at the 5% level, with a coefficient (1.24) slightly smaller than that for the whole sample (1.36).
  - Conclusion: key findings are not driven only by U.S. samples.

### Quantitative magnitudes and broader conclusions (from VI. CONCLUSIONS)
- Sample covered: 11,106 non-financial firms in 16 advanced countries from 2007 to 2013.
- Two bank capital measures examined: the Tier 1 ratio and a simple leverage ratio.
- Core empirical result:
  - Firms with larger external financing needs invest relatively more in countries that are able to maintain higher bank leverage ratios.
  - The effect is particularly stronger for non-financial firms with sound fundamentals.
- Magnitude illustrations:
  - Going from the average level of capital in U.S. banks in 2007 to the level that prevailed in 2013, holding all else constant, had the effect of adding around 0.56 percent of GDP to corporate investment.
  - If the U.S. were today to have the average level of Tier 1 capital-to-total assets prevailing among European banks, it would result in 0.73 percent of GDP lower investment, purely from the effect that lower bank capital has on the availability of credit to firms reliant on external financing.

### Policy-relevant implications
- Countries proactive in addressing banking sector capital shortfalls and providing regulatory and other incentives for banks to raise capital (in the sense of a higher leverage ratio) see dividends in the form of higher corporate investment ratios.
- Firms with sound fundamentals and solid investment projects are less likely to be credit-constrained when bank balance sheets are stronger.

*Source: _wp15145 - 5.2 Sector level regression*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15145.pdf_