## wpiea2025115-print-pdf

## Source details

**Canonical URL:** [wpiea2025115-print-pdf](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025115-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025115-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025115-print-pdf.pdf.json)

---

### Research question and approach
- Main question: Why do European firms predominantly use bank funding while U.S. firms choose more bond issuance? If European firms faced a financial market akin to the U.S. market, would aggregate debt funding choice be the same?
- Approach:
  - Compile a micro-level data set combining bond issuances from Thomson Reuters / Refinitiv and firm data from Bureau van Dijk’s Orbis database.
  - Use an adapted Becker and Josephson (2016) model with firm heterogeneity along firm size and fixed asset share to estimate contributions of firm fundamentals versus financial market structure.
  - Present counterfactual scenarios relocating European firms into a U.S.-like financial market.

### Stylized empirical facts (cross-sectional patterns)
- Aggregate debt funding split (nonfinancial corporations):
  - Europe: around 85%/15% (bank loans/corporate bonds).
  - United States: 45%/55% (bank loans/corporate bonds).
- Distributional patterns:
  - Large firms (frequently public) carry higher share of market (bond) debt; in the United States the bond debt share for this group is almost 70 percent.
  - In the euro area, even the largest quartile of public firms has bond funding share below 50 percent.
  - SMEs in the United States: bank loans make up more than 90 percent of all debt (market debt share below 10 percent).
- Size and market fragmentation:
  - In the United States, 6 out of 10 employees work in a firm with at least 300 employees; only a third of European employees do.
  - Small firms play a more prominent role in the euro area.
- Additional micro-data stylized facts:
  - (i) Firm size is an essential predictor of bond issuance.
  - (ii) Among very large entities, European firms have a smaller share of bond debt compared to U.S. counterparts.
  - (iii) The smaller European bond share holds across smaller size categories.
  - (iv) Cut-off firm size to begin issuing bonds:
    - Smallest American firms issuing bonds employ between 100 and 250 employees.
    - In Europe, bond issuance is almost exclusively seen among firms with at least 5,000 employees.
  - Firms with more redeployable (fixed) assets tend to issue more bonds in both regions.

### Data, coverage, and representativeness
- Bond issuance data:
  - More than 300,000 bond issuances by nonfinancial issuers downloaded from Refinitiv; instruments with equity- or option-like features removed.
  - Remaining outstanding amount determined at each balance sheet cut-off, accounting for buybacks and reopenings.
- Firm balance sheet data:
  - Orbis used for firm-level balance sheets and ownership; consolidated accounts used when available; majority-owned subsidiary accounts dropped to avoid double-counting.
- Sample period and coverage:
  - Data cleaned starts in 2012 and ends just before Covid (dataset cut in 2019).
  - Euro area sample represents about 60 percent of aggregate revenues and 59 percent of total employment.
  - For the United States, counterfactual analysis relies on characteristics of bond-issuing firms; total covered bond debt outstanding represents more than two-thirds of the aggregate in both regions.
- Micro-level representation of bond funding share in 2018:
  - Euro area: 17 percent (micro-level); 13 percent in the aggregate.
  - United States: 53 percent (micro-level); 62 percent in the aggregate.
- Country-level cleaned sample sizes for euro area (final cleaned full sample contains information on):
  - Austria: 41.069 firms
  - Belgium: 64.858 firms
  - Finland: 136.199 firms
  - France: 734.663 firms
  - Germany: 242.772 firms
  - Netherlands: 15.792 firms
- Bond-issuer counts (cleaned dataset):
  - Euro area: 885 bond issuers.
  - United States: 1.511 bond issuers.
  - Among European bond issuers, 48% are listed firms; in the United States 63% are listed firms.
- Coverage metrics for Orbis (aggregate representativeness):
  - Total revenue coverage (euro area): 60% on average.
  - Total number of employees coverage (euro area): 59% on average.
  - Aggregate total operating surplus coverage (euro area): 69% on average.
- Sample bias:
  - Micro-level data on average covers more than two-thirds of aggregate bond debt.
  - The sample slightly overestimates bond funding share in the euro area and modestly underestimates it in the United States.

### Mechanisms and theoretical interpretation
- Key firm characteristics emphasized:
  - Firm size (interpreted as structural, stemming from market fragmentation).
  - Liquidation value proxied by fixed asset share (asset tangibility / collateral availability).
- Theoretical tensions:
  - Asymmetric-information / bank monitoring view: tangible assets reduce information asymmetry and favor bond issuance.
  - Bank-efficient-liquidation view: efficient bank-led out-of-court restructuring can make bank loans relatively more attractive when fixed assets aid recovery.
- Model summary (adapted Becker and Josephson (2016)):
  - Heterogeneous firms differ in project size D (or K) and fixed asset share θ.
  - Project success with probability 1 − q; default with probability q.
  - Bond investors set break-even r* = q(β + σ)/(1 − q).
  - Banks face convex loan cost C(Li) = c Li^2 / 2 and compete à la Cournot with n banks.
  - Out-of-court restructuring yields (1 − β)K; in-court bankruptcy yields (1 − β − σ)K with σ = α − τ · θ.
  - Equilibrium implications:
    - Two regimes: (i) bank-only funding for small firms; (ii) mixed funding with bond marginal for large firms.
    - Bond share increases in firm size D, in fixed asset share θ, and in bank cost-of-capital c; decreases in q and in banking competition n.
    - Cut-off size for bond market entry: D_CutOff = q (n − 1) (α − τ·θ) / c.
    - Higher θ reduces D_CutOff (∂L_CutOff/∂θ = − q τ (n−1) / c).

### Calibration, estimation, and parameters
- Default probability specification: q_i(D_i, θ_i) = q0 + τ_D D_i + τ_θ θ_i.
- Regression-based parameter choices (using 812 firms’ CDS-based default info and assumed recovery rates 68.1 Cts/$ for euro area and 67.5 Cts/$ for the United States):
  - τ_D = −0.003
  - τ_θ = −0.022
- Estimated/calibrated parameters (minimizing squared deviations of model predictions):
  - United States: c = 0.077 ∗∗∗, τ = 0.395 ∗∗∗.
  - Euro area: c = 0.083 ∗∗∗, τ = −0.058.
  - Scaling parameters set to n = 10 and α = .5 in both regions.
- Note: higher cost-of-capital parameter c for Europe consistent with observed higher funding costs of European banks.

### Counterfactual analysis and quantitative findings
- Counterfactual design: estimate two parameter sets (euro area and United States) and simulate intensive and extensive margin responses when relocating euro area firms to a U.S.-like market.
- Key quantitative results:
  - Simple statement in Introduction: If European firms faced a financial market like the United States, their aggregate bond funding share would on average be about 18 percentage points higher (closing less than half the gap).
  - Estimated intensive-margin effect (European bond issuers face U.S. market conditions; cut-off constant):
    - Result: European bond issuers would issue 10.5 percentage points more bonds than currently.
    - These different funding conditions explain around a quarter of the observed difference in bond funding share between regions.
  - Estimated extensive-margin effect (additional firms enter bond market via cut-off shift):
    - Result: Additional entrants increase aggregate bond funding share by, on average, less than 5 percentage points.
    - Implies only a small number of current nonissuers would begin issuing bonds under U.S.-like conditions.
  - Combined intensive + extensive margin:
    - Average bond share for the euro area counterfactual (intensive + extensive margin) is 25.8%.
    - Observed average bond share over 2011–2019: United States = 58.4%, euro area = 13.1% (average difference = 45.3%).
    - Counterfactual intensive + extensive margins leave a difference of 32.6 percentage points between observed U.S. firms and euro area counterfactual.
- Attribution of modeled variation:
  - Different firm fundamentals explain two-thirds of the modeled variation in bond funding share between regions.
  - Differences in financial market structure explain the remaining one-third.

### Empirical regularities emphasized for identification
- Regression evidence (motivational regressions):
  - Univariate regressions:
    - Firm size measured by total assets: R2 = 12.9 %
    - Firm size measured by number of employees: R2 = 6.7 %
    - Profitability and leverage: R2s of negligible size
  - Conclusion: firm size and fixed assets (asset tangibility) are the most important drivers of debt choice in the sample.
- Firm-size distribution effects (back-of-the-envelope calibrations):
  - Euro area bond funding share would increase from 13 percent to 25 percent if it had the U.S. firm size distribution (all else equal).
  - United States bond funding share would be 31 instead of 55 percent if it had the euro area firm distribution.
- Employment-weighted importance:
  - Firms with less than 300 employees rarely issue bonds in either region.
  - In the euro area these small firms account for 62% of total employment, compared to 43% in the United States.

### Policy implications and contributions
- Contributions:
  - First paper using aggregate micro-level data to present counterfactual debt scenarios for the euro area.
  - Improves coverage of private and small firms relative to Capital IQ–based studies by combining Refinitiv bond data and Orbis balance sheets.
- Policy relevance:
  - Differences in aggregate funding choices and associated systemic risk in the euro area are substantially driven by structurally different firm characteristics in a segmented market, not solely by financial systems.
  - Interventions should balance measures to foster debt market access for small firms and measures to disincentivize banks’ amplifying behavior, acknowledging that reliance on bank debt may be unavoidable given the European firm distribution.
  - Long-term: reducing market fragmentation could change firm distribution and thereby impact aggregate funding choice more strongly.
  - EU capital markets union objectives are informed by these findings but short- to medium-term constraints from firm distribution limit bond uptake.

### Appendix highlights (data cleaning and firm distribution)
- Data cleaning rules included dropping observations with unknown or negative equity, unknown or zero operating revenue, fixed asset share < 0 or > 1, bond funding share < 0 or > 1, revenues > 500 million and employees < 5, and removing financial companies and certain NACE categories.
- Appendix IV (firm distribution):
  - Figure weights firms by number of employees; census sources used include Statistics of U.S. Businesses (SUSB) and Eurostat structural business statistics.
  - Reconciliation method for inconsistent census buckets: example calculation yields weight for size category 50-299 employees = 0.1824 (18.2%).
- Summary statistics and country-level tables presented for employment, revenue, total assets, and outside funding for Austria, Belgium, Germany, Finland, France, Netherlands and bond issuers subset.

*Source: wpiea2025115-print-pdf - Introduction and selected sections*

### Introduction3

### Introduction3

### Data8
- Data8

### Debt choice and a heterogeneous cross-section12
- Debt choice and a heterogeneous cross-section12
- Firm characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  14
- Modeling choice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  21

### Counterfactual analysis21
- Counterfactual analysis21
- Model description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  21
- Equilibrium with bond and bank debt . . . . . . . . . . . . . . . . . . . . . . . . . .  24
- Equilibrium with bank debt only . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  25
- Calibration and estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  26
- Counterfactual scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  27

### Conclusion30
- Conclusion30

### Appendix31
- Appendix31
- Appendix I: Data Cleaning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  31
- Appendix II: Summary Satistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  32
- Appendix III: Firm coverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  34
- Appendix IV: Firm distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  37
- Appendix V: Information on the distribution of bond funding in the economy . . . .  38

### References43
- References43

*IMF WORKING PAPERSNatural Bank Reliance*

### Introduction

### Introduction

### Research question and approach
- Main question: Why do European firms predominantly use bank funding while U.S. firms choose more bond issuance? If European firms faced a financial market akin to the U.S. market, would aggregate debt funding choice be the same?
- Approach:
  - Compile a unique, extensive micro-level data set combining bond issuances from Thomson Reuters / Refinitiv and firm data from Bureau van Dijk’s Orbis database.
  - Use a theoretical model (adapted from Becker and Josephson (2016)) with firm heterogeneity along two dimensions—firm size and fixed asset share—to estimate the contribution of firm fundamentals versus financial market structure to aggregate debt choice.
  - Present counterfactual scenarios relocating European firms into a U.S.-like financial market.

### Stylized empirical facts (cross-sectional patterns)
- Aggregate debt funding split (nonfinancial corporations):
  - Europe: around 85%/15% (bank loans/corporate bonds) in the aggregate.
  - United States: 45%/55% (bank loans/corporate bonds) in the aggregate.
- Distributional patterns common to both regions:
  - Large firms (frequently public) carry a higher share of market debt (bond debt). In the United States, the bond debt share in this group is almost 70 percent.
  - In the euro area, larger firms also employ more bond debt, but the bond funding share remains below 50 percent even for the largest quartile of public firms.
  - Firms at the bottom of the firm size distribution almost exclusively use bank loans. Among SMEs in the United States, bank loans make up more than 90 percent of all debt, implying a share of market debt below 10 percent.
- Size and market fragmentation:
  - In the United States, 6 out of 10 employees work in a firm with at least 300 employees; only a third of European employees do.
  - Small firms play a more prominent role in the euro area than in the United States.
- Additional stylized facts from the micro data:
  - (i) Firm size is an essential predictor of bond issuance.
  - (ii) Among very large entities, European firms have a smaller share of bond debt compared to U.S. counterparts.
  - (iii) The observation that European firms hold smaller shares of bond debt also holds in smaller size categories.
  - (iv) The cut-off firm size to begin issuing bonds is higher in Europe than in the United States:
    - Smallest American firms issuing bonds employ between 100 and 250 employees.
    - In Europe, bond issuance is almost exclusively seen among firms with at least 5,000 employees.
  - Firms with more redeployable (fixed) assets tend to issue more bonds in both regions. Collateral referred to as fixed assets or redeployable assets does not necessarily imply these assets are pledged.

### Mechanisms and theoretical interpretation
- Two firm characteristics emphasized:
  - Firm size (interpreted as structural, stemming from market fragmentation and not driven by funding choice).
  - Liquidation value proxied by fixed asset share (asset tangibility / collateral availability).
- Theoretical implications:
  - Models based on asymmetric information and improved monitoring by banks suggest tangible assets reduce information asymmetry and benefit bond issuance.
  - Models based on more efficient liquidation via banks suggest a large share of tangible assets benefits choosing bank loans.
- Model used:
  - Adapted version of Becker and Josephson (2016) incorporating heterogeneity in firm size and fixed asset share.
  - Firms undertake a project with exogenous size and fixed asset requirement.
  - Interaction in Cournot competition yields an equilibrium similar to data: only firms beyond a size threshold issue bonds, and this threshold increases with a firm’s fixed asset share.
  - Mechanism: banks are more efficient at recovering value via out-of-court restructuring; bond holders rely on formal bankruptcy procedures which destroy more value when redeployable assets are scarce, making bond investment riskier.

### Counterfactual scenarios and quantitative findings
- Estimation:
  - Two parameter sets estimated separately to represent euro area and United States debt choice patterns.
  - Counterfactuals use estimated parameters and aggregate firm distributions to simulate intensive and extensive margin changes in bond funding.
- Key quantitative counterfactual result:
  - If European firms faced a financial market like the United States, their aggregate bond funding share would on average be about 18 percentage points higher, closing less than half of the gap between the two countries.
  - This increase is primarily driven by existing bond issuers issuing more bonds ("counterfactual constant set" of issuers); the additional entry of smaller firms beginning to issue bonds ("counterfactual additional issuers") increases the aggregate share only slightly.
- Attribution of modeled variation:
  - Different firm fundamentals explain two-thirds of the modeled variation in bond funding share between regions.
  - Differences in the financial market structure explain the remaining one-third.

### Data, coverage, and sample
- Bond issuance data:
  - Extracted from Refinitiv (formerly Thomson Reuters) debt deals database, covering all historic bond issues by nonbank corporates, including Bonds, Notes, Debentures, Commercial Paper, Negotiable European Commercial Paper (Short-Term or Medium Term) as defined under French law, Inhaberschuldverschreibungen (specific German bond), and certain notes and types of commercial paper.
  - Instruments with equity- or option-like features (such as convertible bonds or warrants) were removed.
  - Information on more than 300,000 bond issuances by nonfinancial issuers downloaded.
  - Remaining outstanding amount for each bond determined at each balance sheet cut-off date, taking buybacks and reopenings into account.
- Firm balance sheet data:
  - Bureau van Dijk’s Orbis database used for firm-level balance sheets and ownership information; consolidated accounts used when available; majority-owned subsidiary accounts dropped annually to avoid double-counting.
- Combined micro-level data set:
  - After data cleaning, the data starts in 2012 and ends just before Covid.
  - Covers private and public nonfinancial firms in the euro area and the United States.
  - The euro area sample represents about 60 percent of aggregate revenues and 59 percent of total employment.
  - For the United States, the counterfactual analysis relies on characteristics of bond-issuing firms; total covered bond debt outstanding represents more than two-thirds of the aggregate in both regions.
  - The share of bond funding (over total debt funding) represented in micro-level data in 2018:
    - Euro area: 17 percent (micro-level); 13 percent in the aggregate.
    - United States: 53 percent (micro-level); 62 percent in the aggregate.
  - Final cleaned full sample for the euro area contains information on:
    - 41.069 firms from Austria
    - 64.858 firms from Belgium
    - 136.199 firms from Finland
    - 734.663 firms from France

### Policy implications and contributions
- Contributions:
  - First paper using aggregate micro-level data to present counterfactual debt scenarios for the euro area.
  - Contributes to macro-financial literature on debt markets and empirical literature on debt choices among heterogeneous firms.
  - Improves coverage of private and small firms relative to Capital IQ–based studies by combining Refinitiv bond data and Orbis balance sheets.
- Policy relevance:
  - Findings inform the EU capital markets union debate aiming to promote more efficient funding choices and improve integration and resilience of financial markets.
  - Results suggest that differences in aggregate funding choices and associated systemic risk in the euro area are not solely due to financial systems but are substantially driven by structurally different firm characteristics in a segmented market.
  - Policy implication: interventions should balance measures to foster debt market access for small firms and measures to disincentivize banks’ amplifying behavior, acknowledging that reliance on bank debt may be unavoidable given the European firm distribution.
  - Long-term: less market fragmentation could change the firm distribution and thereby impact aggregate funding choice more strongly.

*Source: wpiea2025115-print-pdf - Introduction*

### 242.772 firms from Germany and 15.792 firms from the Netherlands. I focus on the countries less af-

### Natural Bank Reliance

### Sample and data cleaning
- The micro-level firm sample includes "242.772 firms from Germany and 15.792 firms from the Netherlands."
- The dataset is cut in 2019 to avoid confounding effects of the Covid pandemic.
- The cleaned dataset on bond issuers contains information on 885 bond issuers in the euro area and 1.511 bond issuers in the United States.
- Among the European bond issuers, 48% are listed firms, compared to 63% in the United States.
- The analysis focuses on countries less affected by the sovereign debt crisis (also referred to as Non-GIIPS countries) to ensure that results are not driven by that crisis.

### Coverage and representativeness of Orbis firm data
- Aggregate measures considered to assess representativeness:
  - Total revenue of nonfinancial corporates: coverage for the euro area is, on average, 60% of revenue covered.
  - Total number of employees in the nonfinancial corporate sector: average coverage in covered countries of the euro area is 59%.
  - Aggregate total operating surplus: Orbis data covers about 69% of the aggregate.
- If corporate groups report worldwide employment, this is broken down to the domestic level by the share of local revenue in worldwide revenue.
- A set of summary statistics for all firms and for bond issuers in each country are reported in the appendix (Table 2 in the source).

### Bond data versus aggregate debt securities
- Figure 2 compares total volume of outstanding bond debt across the cleaned micro-level data to aggregate data for two regions: the euro area (Non-GIIPS) and the United States.
  - Aggregate outstanding volumes of debt securities are depicted in blue.
  - Micro-level data are depicted in orange (raw bond data, according to the borrower country reported by Refinitiv) and yellow (firm-matched data, by the firm’s domicile in Orbis).
- The micro-level data set, on average, covers more than two-thirds of the aggregate bond debt.
- The aggregate data for the euro area was derived using the time series Total Debt Securities for all core European countries from the quarterly financial and nonfinancial sector accounts available in the ECB Statistical Data Warehouse.
- The aggregate data for the United States was derived from the time series Nonfinancial Corporate Bonds (CBLBSNNCB) obtained from Federal Reserve Economic Data.
- Nuanced differences in the definition of domicile explain the minor overshooting in the EA coverage.

### Definition and computation of bond funding share
- Bond Funding Share is computed as:
  Bond Funding Share = Bond Debt Outstanding / (Total Debt − Nonfinancial Debt)
- The bond funding share refers to the share of bond funding in external financial funding.
- The volume of outstanding financial debt is computed as the difference between total debt and other nonfinancial debt items.
  - These nonfinancial debt items include other current liabilities (creditors) and other non-current liabilities (provisions).

### Aggregate bond funding share trends and sample bias
- Figure 3 plots the aggregate bond funding share across all covered firms over time and compares the micro-level dataset observation to macroeconomic observations reported by Eurostat and the Federal Reserve.
- United States:
  - Bond share increased slightly from around 55% in 2011 to more than 60% in 2018.
- Covered countries of the euro area:
  - Bond funding share increased slightly from approximately 11% to around 13%.
- Sample bias observations:
  - The sample slightly overestimates the bond funding share in the euro area.
  - The sample modestly underestimates the bond funding share in the United States.
- For the counterfactual analysis:
  - Detailed coverage for the euro area is required.
  - For the United States, only good coverage among the bond-issuing firms is necessary, since no consolidated nationwide micro-level coverage is required for that region.

*Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025115-print-pdf.pdf*

### conclusions are to be drawn about nonissuers in the United States and these are not incorporated into

### wpiea2025115-print-pdf - conclusions are to be drawn about nonissuers in the United States and these are not incorporated into

### Data coverage and measurement
- Total coverage of U.S. firms in Orbis is smaller than for Europe; coverage of bond-issuing firms is very good in both regions (according to Figure 2).
- Aggregate series used to construct region-level bond shares:
  - Euro area: "Total Debt Securities And Loans" and "Total Debt Securities" from ECB Statistical Data Warehouse.
  - United States: "Nonfinancial Corporate Debt Securities Total Liabilities" (NCBDBIQ027S) and "Nonfinancial Corporate Business Loans" (NCBLL) from Federal Reserve Economic Data.
- Micro-level aggregate constructed from firm-level micro data as detailed in the paper.

### Connection between micro and macro literatures on debt choice
- Two strands of literature:
  - Corporate finance: microeconomic choice of individual firms between bank loans and bonds.
  - Macroeconomic: aggregate volumes of debt and role of institutional/political factors.
- Institutional factors influencing debt choice highlighted in the literature include: efficiency of bankruptcy procedures, power of majority shareholders, financial reporting quality.
- Key interpretation objective: determine whether aggregate differences in bond-vs-loan funding between the euro area and the United States arise because (a) firms with similar characteristics make different choices across regions (potential inefficiency/bias), or (b) structurally different firm populations lead to different aggregate outcomes.

### Firm heterogeneity and empirical patterns
- Firm size
  - Large firms are more likely to issue bonds (cited micro literature).
  - Regional firm-size distribution differences:
    - Europe: between 45% and 50% of employees work in firms with less than 49 co-workers.
    - United States: only about a quarter of employees work in firms with less than 49 co-workers.
  - Cut-offs and bond prevalence:
    - United States: bond funding becomes prevalent at around 300 employees.
    - Euro area: smallest bond-issuing firms tend to have more than 5,000 employees.
    - Among largest firms: bond funding share grows to above 50% in the United States, while in the euro area it remains below 30%.
  - Aggregate counterfactuals (back-of-the-envelope calibration):
    - Euro area bond funding share would increase from 13 percent to 25 percent if it had the U.S. firm size distribution (all else equal).
    - United States bond funding share would be 31 instead of 55 percent if it had the euro area firm distribution.
  - Employment-weighted importance:
    - Firms with less than 300 employees rarely issue bonds in either region.
    - In the euro area these small firms account for 62% of total employment, compared to 43% in the United States.
- Financial constraints and firm-size distribution (FSD)
  - SAFE survey: access to finance reported as major obstacle by 15.4% of SMEs in 2010 and below 8% in 2019.
  - Other SME obstacles: customers 27.8% (2010) / 20.9% (2019); skilled employees 12.4%/24.2%.
  - EIB Investment Survey: 7.19 percent of EU SMEs experienced financial constraints in 2023.
  - Prior literature (Angelini and Generale (2008); Beck et al. (2006)) suggests funding constraints play a limited role in shaping FSD across developed economies.
- Asset tangibility / liquidation value
  - Firms with higher fixed-asset shares (greater asset tangibility) use more bond funding.
    - United States: bonds about 20% of financial funding for firms with small fixed-asset share (< one quarter of total assets) and more than 50% for firms with large fixed-asset shares (two-thirds of total assets and above).
    - Euro area: bond funding also increases with fixed assets but remains below 20% for firms with large fixed assets; slope lower than in the U.S.
  - Asset tangibility differs strongly by industry and is treated as exogenous in the modeling exercise.
- Leverage and profitability
  - Leverage: depicted in appendix; not identified as the main driver of debt choice in the sample.
  - Profitability (ROA measured with EBIT): no clear relation between bond funding and profitability across regions.
- Regression evidence (motivational regressions; Table 3 in appendix)
  - Univariate regressions:
    - Firm size measured by total assets: R2 = 12.9 %
    - Firm size measured by number of employees: R2 = 6.7%
    - Profitability and leverage: R2s of negligible size
  - Conclusion: firm size and fixed assets (asset tangibility) are the most important drivers of debt choice in the sample; profitability and leverage are less important.

### Modeling choice and identification strategy
- Modeling aim: compute counterfactuals that separate effects of firm characteristics (heterogeneous cross-section) from institutional/legal differences.
- Modeling assumptions and scope:
  - Partial-equilibrium model of funding choice with heterogeneous firms differing in project size and share of fixed assets (θ).
  - Firm capital demand: K(r) = A − B r with inverse R(K) = (A − K)/B.
  - Project success probability: success with probability 1 − q; default with probability q.
  - Investors:
    - Atomistic bond investors and nbanks; risk-neutral.
    - Main specification: bond and bank debt rank pari-passu.
    - Atomistic bond holders never engage in out-of-court restructurings and fully rely on in-court procedures.
  - Bankruptcy / restructuring:
    - Out-of-court restructuring yields firm value (1 − β) K(r).
    - Formal in-court bankruptcy yields firm value (1 − β − σ) K(r), with β, σ > 0 and (1 − β − σ) > 0.
    - Welfare loss σ depends on firm’s share of fixed assets: σ = α − τ · θ (welfare loss decreases with higher θ).
  - Bond supply
    - Bond investors require return r* and supply is perfectly elastic at r*.
    - Closed-form: r* = q(β + σ)/(1 − q).
  - Banks and lending costs
    - Banks subject to regulation/capital requirements; convex loan cost function:
      - C(Li) = c L^2_i / 2
    - Banks compete à la Cournot; when market interest rate < r* banks supply all funding; at r* bond investors provide infinite funding at that break-even rate and bond funding enters equilibrium.
  - Institutional differences
    - Model allows euro area and U.S. market environments to differ in restructuring efficiency and balance-sheet cost determinants.
    - Legal-system differences (common law vs civil law) captured through parameters α and τ affecting σ.
- Equilibrium implications (qualitative)
  - Two equilibria can arise (for small and large firms) under reasonable parameter values; which equilibrium materializes depends mainly on firm size.
  - Banks can internalize welfare gains from out-of-court restructuring and match bond holders’ in-court returns, distributing σ·K among themselves.

### Counterfactual analysis approach
- Use adapted Becker and Josephson (2016) model extended for heterogeneity in fixed-asset shares (θ) and firm sizes to compute counterfactuals (e.g., how euro area bond share would change under U.S. firm-size distribution).
- Key identification: treat firm size and asset tangibility as exogenous characteristics shaping funding choice; incorporate insolvency procedure efficiency as an institutional parameter.

*Source: IMF Working Paper (excerpts from the provided PDF content).*

### 1. Large firms: both banks and bond holders provide funding. Bond holders are the marginal in-

### 1. Large firms: both banks and bond holders provide funding. Bond holders are the marginal in-

### Scenarios and stylized setup
- Scenario (1): Large firms obtain funding from both banks and bond holders; bond holders are the marginal investors and the interest rate is r*; the equilibrium quantity of debt demanded by those firms at the margin is D.
- Scenario (2): Smaller firms rely solely on bank loans at rates below r*; the equilibrium quantity is smaller than D; bond holders do not lend to firms in this size category.
- Key modeled drivers of the funding choice: firm size (D), fixed asset share (θ), bank cost of capital (c), probability of bad project outcomes (q), and banking sector competition (n).

### Equilibrium with bond and bank debt
- Individual bank profit-maximization problem (symmetric banks, loan by bank i denoted Li, other banks L−i):
  - U(Li, L−i) = (1−q)∗R(K)∗Li + q∗[σ∗K∗ (Li L− (β+σ)Li)] − c Li^2/2. (equation 6)
- In the good outcome, loan Li is fully repaid with interest; in the bad outcome, out-of-court restructuring yields return equal to in-court bankruptcy payout plus welfare gain σ split among banks; capital cost term c Li^2/2 applies regardless of outcome.
- If bond funding is provided: R(K) = r* = q(β+σ)/(1−q). Total funding demanded at this rate is D = K(r*).
- Individual bank first-order condition:
  - ∂U(Li, L−i)/∂Li = qKσ (L−Li)/L^2 − cLi = 0. (equation 7)
- Optimal total bank loan supply:
  - L = √(q∗σ∗D∗(n−1)/c). (equation 8)
- Optimal bond share (Bond Share (BS)):
  - Bond Share(BS) = 1 − L/D = 1 − √(q∗(n−1)∗(α−τ·θ) D∗c). (equation 9)
- Comparative statics (as stated):
  - Bond share increases in firm size (D) and in the share of fixed assets owned (θ).
  - Bond share increases in the bank’s cost of capital (c).
  - Bond share decreases in the probability of bad project outcomes (q) and in banking competition (n).

### Equilibrium with bank debt only (cut-off)
- Bank loans suffice at the cut-off if:
  - L = √(q∗(n−1)∗(α−τ·θ)∗D / c) ≥ D. (equation 10)
- Critical threshold (cut-off size) at which firms begin to receive bond funding:
  - D_CutOff = L_CutOff = q∗(n−1)∗(α−τ·θ) / c. (equation 11)
- Comparative statics for the cut-off:
  - Cut-off increases with q and n.
  - Cut-off decreases with θ and c.
  - Marginal effect of fixed asset share on cut-off:
    - ∂L_CutOff/∂θ = − q τ (n−1) / c. (equation 12)
- Economic rationale: larger θ reduces welfare losses to bond holders because fixed assets are easier to recover in bankruptcy; hence bond investors find lending more attractive for firms with higher θ.

### Model features and empirical stylized facts accommodated
- Model accommodates mixed debt choices (firms can use both banks and bonds), in contrast to single-source models.
- Determinants of cut-off to enter bond market: fixed assets (asset tangibility) and firm size, not leverage, risk, or profitability.

### Calibration and estimation
- Default probability modeled as q_i(D_i, θ_i) = q0 + τ_D D_i + τ_θ θ_i; literature suggests default probability decreases with firm size and varies by industry.
- Using 812 firms’ CDS-based default information and assumed recovery rates of 68.1 Cts/$ for the euro area and 67.5 Cts/$ for the United States, the regression yields:
  - τ_D = −0.003 and τ_θ = −0.022 (set based on regression results).
- Data aggregation: yearly distribution of bond funding share across 9 (size) × 10 (fixed asset share) categories → 90 data points per year per region (r = US/EZ). Cut-off defined at bond share of 1%.
- Estimated/calibrated parameters (minimizing squared deviations of model predictions for the cut-off size and bond share):
  - United States: c = 0.077 ∗∗∗, τ = 0.395 ∗∗∗.
  - Euro area: c = 0.083 ∗∗∗, τ = −0.058.
- Scaling parameters set to n = 10 and α = .5 in both regions.
- Note: higher model cost-of-capital parameter for Europe aligns with observed higher funding costs of European banks.

### Counterfactual scenarios and quantitative findings
- Intensive margin counterfactual (European bond issuers face U.S. market conditions; firm cut-off held constant):
  - Result: European bond issuers would issue 10.5 percentage points more bonds than currently.
  - These different funding conditions explain around a quarter of the observed difference in bond funding share between the two regions.
- Extensive margin counterfactual (allow additional firms to enter bond market via shift in cut-off):
  - Result: Additional entrants increase aggregate bond funding share by, on average, less than 5 percentage points.
  - Implies only a small number of current nonissuers would begin issuing bonds under U.S.-like conditions.
- Remaining difference and attribution:
  - Observed average bond share over 2011–2019: United States = 58.4%, euro area = 13.1% (average difference = 45.3%).
  - Counterfactual intensive + extensive margins leave a difference of 32.6 percentage points between observed U.S. firms and euro area counterfactual.
  - Interpretation: more than two-thirds of the observed funding difference can be attributed to different firm characteristics; roughly one-third is associated with different market conditions.
  - The average bond share for the euro area counterfactual analysis (intensive + extensive margin) is 25.8%.

### Conclusion — substantive implications and policy relevance
- The greater reliance of European firms on bank funding relative to U.S. firms is largely explained by:
  - (i) European firms being smaller on average; small firms rely more on bank funding.
  - (ii) The size cut-off for bond funding being considerably higher in the United States.
  - (iii) European firms with large fixed asset shares tending to rely more on bank debt than U.S. counterparts.
- Quantitative takeaway: about two-thirds of the funding difference is due to firm characteristics; about one-third due to market conditions.
- Policy implications:
  - Market fragmentation and firm-size distributions are central to understanding Europe’s “natural” bank reliance.
  - The European Union’s capital markets union project aims to promote more efficient funding choices; however, short- to medium-term constraints from firm distribution limit bond uptake.
  - Long-run changes in capital market structure and reductions in market segmentation can alter aggregate funding choices endogenously via firm distribution shifts.
  - Policymakers should factor in Europe’s natural reliance on bank funding when designing measures to change the financial market structure and adoption speed.

### Appendix highlights (data cleaning and coverage)
- Data cleaning rules included dropping observations with:
  - unknown or negative equity (‘Shareholders Funds’), unknown or zero operating revenue (‘Operating Revenue / Turnover’);
  - fixed asset share < 0 or > 1; bond funding share < 0 or > 1; revenues > 500 million and employees < 5.
- Ensured domestic non-financial corporate coverage by removing:
  - Entity type = ’Financial company’ or ’Insurance company’;
  - unknown NACE (Rev.2) main section; NACE core code starting with 64; NACE main section = ’U - Activities of extraterritorial organisations and bodies’ or ’T - Activities of households as employers; undifferentiated goods- and services-producing activities of households for own use’.
- Summary statistics and country-level tables presented for firm employment, revenue, total assets, and outside funding (means and standard deviations) for Austria, Belgium, Germany, Finland, France, Netherlands and for bond issuers subset.

*International Monetary Fund — IMF Working Paper: Natural Bank Reliance*

### Appendix IV: Firm distribution

### Appendix IV: Firm distribution

### Figure 4: Cumulative firm size distribution — interpretation and data sources
- The Figure weights firms by the number of employees; an observation indicates how many employees work in firms smaller than a given cut-off.
- The majority of data points in this Figure draw on census data and are therefore not prone to selection bias.
- Census sources referenced:
  - Statistics of U.S. Businesses (SUSB) of the U.S. Census Bureau.
  - Statistics on small and medium-sized enterprises from Eurostat (Eurostat’s structural business statistics).

### Method for reconciling U.S. and European census bucket inconsistencies using Orbis
- The U.S. and European census on firm size do not use consistent bucket sizes; Orbis is used to adjust buckets.
- European census groups all firms of at least 250 employees in one bucket, representing 38.55% of all employees.
- To deduce the share of employees in the European group 250-299:
  - Compute the share of the size group 250-299 in terms of all firms of at least 250 employees across nonfinancial Euroarea firms in Orbis (Orbis Pivot analysis tool), which is 2.212%.
  - Deduce that 0.85% of employees work in firms of the size category 250-299 employees using: 0.02212 * 0.3855 = 0.0085 (0.85%).
- The European census already includes the share of all employees in the size group 50-249 employees, which is 17.39%.
- Resulting weight for the size category 50-299 employees as depicted in the graph:
  - 0.1739 + 0.0085 = 0.1824, or 18.2%.

### Data representativeness and smaller size categories
- Orbis data has been found to be representative in terms of employment for firms of at least 250 employees; see (Bajgar et al., 2020, p. 22).
- The three smallest size categories (<10, 10-19, 20-49), which are most prone to selection biases, are available as is in both census publications.

*Source: wpiea2025115-print-pdf - Appendix IV: Firm distribution*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025115-print-pdf.pdf_