## wpiea2020294-print-pdf

## Source details

**Canonical URL:** [wpiea2020294-print-pdf](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020294-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020294-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020294-print-pdf.pdf.json)

---

### Introduction and research questions
- Investigates whether the post-Global Financial Crisis (GFC) productivity slowdown impacted firms unevenly across the size distribution and why.
- Two main questions:
  - Did smaller firms experience a larger decline in post-crisis TFP growth vis-à-vis their larger counterparts?
  - Can any observed productivity slowdown for smaller firms be explained by ex-ante limited credit market access?
- Firm size definitions (Eurostat): micro ≤ 10 employees; small ≤ 50 employees; medium ≤ 250 employees; large ≥ 250 employees.
- Final sample covers 8 European countries in the period before and after the GFC.
- Key event used to identify credit supply tightening: collapse of Lehman Brothers on September 15, 2008.

### Data and measurement
- Data sources and construction:
  - Merged Orbis, AMADEUS, and Fitch Connect (Bankscope) to build cross-country firm-level dataset.
  - Dataset includes firm-level inputs, outputs, balance sheets, each firm’s main bank(s), bank balance sheets, and for a subset, PATSTAT patent applications.
  - BANKER variable lists up to five creditor banks per firm; 74 percent of bank names matched via probabilistic record linkage.
  - Final sample countries: Denmark, France, Germany, Netherlands, Portugal, Spain, Sweden, United Kingdom.
  - Sample coverage relative to Eurostat SBS: France, Portugal, Spain > 60 percent; Denmark, Germany, Sweden between 30 and 50 percent; Netherlands between 15 and 20 percent.
  - Firms kept: continuously existing over sample period (4 years before and after the GFC).
- Measurement choices:
  - TFP estimated via Wooldridge (2009) one-step efficient GMM control function:
    - y_ijct = a_ijct + β_kj × k_ijct + β_lj × l_ijct−1 + ε_ijct
    - TFP: a_ijct = y_ijct − β̂_ks × k_ijct − β̂_ls × l_ijct−1
    - TFP growth: a_ijct − a_ijct−1
  - Credit supply tightening measures:
    - Average CDS spread change of domestic banks between September 7 and September 28, 2008 (∆CDS, basis points, 5-year CDS).
    - Average CDS spread change of firm’s main creditor banks.
    - Bank Lending Survey–based measures (euro area BLS).

### Main empirical findings — firm size premium and magnitudes
- Aggregate descriptive statistics:
  - Average pre-crisis TFP growth: 3 percent.
  - Post-crisis TFP growth: -2 percent (post-crisis drop = 5 percentage points; ∆TFP growth mean = -0.05).
  - Mean employees = 26; median = 7. Mean age = 19 years.
- Benchmark size effects (specification (2), Table 2):
  - ln(Employees) coefficient positive and significant: smaller firms experienced larger drops in post-crisis TFP growth relative to larger firms.
  - Column 3: a firm’s post-crisis TFP growth is on average 2.4 percentage points lower than its counterpart with about 170 percent more employees (one-unit increase in ln employees).
  - One standard deviation decrease in employment (1.27 units of ln employees) accounts for 18 percent of a standard deviation in the drop in post-crisis TFP growth.
  - Micro, small, medium point estimates for declines relative to large firms (Column 6):
    - Medium-sized: 1.4 percentage points lower.
    - Small: 3,1 percentage points lower.
    - Micro: 5.6 percentage points lower.

### Interaction with tightened credit conditions (∆CDS and creditor-bank exposure)
- Country-level ∆CDS interactions (Table 3):
  - ln(Employees) × ∆CDS interaction significantly positive: smaller firms’ TFP drop larger in countries with larger bank CDS increases.
  - Column 3: in a country with average increase in bank CDS spread (21 basis points), a one-unit decrease in ln employees associated with 0.9 percentage point drop in post-crisis TFP.
  - For ∆CDS one-standard deviation higher than average (7.9 basis points higher), corresponding decline for one-unit decrease in ln employees = 2.4 percentage points.
  - For a country with ∆CDS one-standard deviation above average, average decline relative to large firms:
    - Medium-sized: 0.9 percentage points lower.
    - Small: 2.1 percentage points lower.
    - Micro: 2.8 percentage points lower.
- Firm-level creditor-bank ∆CDS (Table 4):
  - Qualitatively similar results with smaller sample.
  - Column 3: one-unit decrease in ln employees associated with 1.3 percentage point drop for firm whose creditor banks had average exposure to Lehman.
  - If creditor banks experienced ∆CDS one standard deviation higher than average, corresponding decline = 3.2 percentage points.
  - Column 6: with same creditor-bank ∆CDS increase, average decline for medium, small, micro relative to large firms = 0.3, 1.9, and 6.1 percentage points respectively.

### Credit market access channel and creditor strength
- Creditor strength measures:
  - Bank capitalization: average pre-crisis regulatory Tier 1 capital (percent of RWA) above median (BankCapital dummy).
  - CDS presence: at least one creditor bank traded in single-name CDS (CDSpresence dummy).
- Bank capitalization results (Table 5):
  - ln(Employees) × BankCapital interaction significantly negative: firm size premium smaller for firms with better-capitalized creditor banks.
  - Column 3: among firms with creditor banks capitalization above median, a one-unit decrease in ln employees associated with 1-percentage-point drop in post-crisis TFP; for below-median creditor banks, corresponding decline = 2.7 percentage points.
- CDS presence results (Table 6):
  - ln(Employees) × CDS presence interaction significantly negative: access to banks with CDS presence reduces firm size premium.
  - Column 3: among firms with creditor banks with CDS presence, one-unit decrease in ln employees associated with 3.3-percentage-point drop; among firms without CDS presence, corresponding decline = 2.6 percentage points.
  - Column 6: access to banks with CDS presence reduces post-crisis TFP decline for medium, small, micro from 1.9 to 1.3, 4.2 to 2.6, and 7 to 5 percentage points respectively.

### Extensive margin of credit market access (likelihoods)
- Larger firms more likely to have relationships with better-capitalized banks and banks with CDS presence:
  - Table 7 column 3: one-unit increase in ln employees associated with 29 percentage points increase in likelihood of access to high-capitalization creditor bank (explains 110 percent of standard deviation in likelihood).
  - Table 7 column 6: micro firms 84 percent lower likelihood of access to better-capitalized banks.
  - Table 8 column 3: one-unit increase in ln employees associated with 7 percentage points increase in likelihood of CDS presence (explains 16 percent of standard deviation).
  - Table 8 column 6: micro, small, medium likelihoods of CDS presence are 49, 45, and 19 percentage points lower respectively.

### Channels — intangible capital and innovation
- Intangible investment responses (Tables 9 and 10):
  - Intangible investment rate = change in stock of intangible assets / value added.
  - Intangible asset share = stock of intangible assets / (physical + intangible assets).
  - Smaller firms cut intangible investment and reduce intangible asset share more than larger firms.
  - One-unit increase in ln employees associated with post-crisis intangible investment rate on average 0.1 percentage points lower and intangible share 0.4 percentage points lower than counterpart with about 170 percent more employees.
  - Intangible investment rate reductions for medium, small, micro relative to large: 0.3, 0.7, and 0.8 percentage points lower respectively.
- Patent applications (Table 11):
  - Smaller firms reduce patent applications more than larger firms.
  - Column 3: one-unit decrease in ln employees associated with reduction of one patent application in post-crisis period.
  - Column 6: micro, small, medium firms reduce patent applications per year by 2.2, 1.9, and 1.8 respectively relative to large firms (sample mean patent applications ≈ 2 per year).

### Causality, timing, placebo tests, and robustness
- Causality and timing arguments:
  - Firm size is highly persistent and predates the GFC (ln employees autocorrelation = 0.95).
  - Smaller firms did not on average experience slower TFP growth than larger firms before the crisis.
  - Controls include industry×country fixed effects, firm-level balance sheet indicators, firm age, and pre-crisis TFP level.
- Placebo test (post-2000 recession):
  - No evidence that the credit access channel operated in the post-2000 recession: TFP growth of smaller firms did not decline more than larger counterparts; results did not depend on creditor bank strength.
- Horserace of size versus balance sheet vulnerabilities (Tables 13 and 14):
  - Firm size effect strengthens when controlling for balance sheet characteristics; balance sheet effects weaken when controlling for firm size.
  - Example: one-unit decrease in ln employees associated with higher post-crisis TFP growth of 0.8 percentage point without controls and 2.4 percentage points with controls (Table 13).
  - Leverage effect diminishes by 32 percent when adding firm size; debt maturity and liquidity effects drop about 90 percent and lose significance when controlling for firm size.
  - Conclusion: firm size is an important and distinct vulnerability indicator.
- Robustness checks summary:
  - Top-bank instead of averaged creditor banks: similar results.
  - Alternative credit measure (euro area BLS, Table A3): interaction of employment and credit tightening significantly positive.
    - BLS Column 3: in a country with average tightening (72 percent of banks tightened), one-unit decrease in ln employees associated with 2.5 percentage point drop in post-crisis TFP; if tightening one-standard deviation higher (84 percent tightened), corresponding decline = 3.0 percentage points.
    - Micro firms exception: effect not statistically significant in one BLS estimate.
  - Alternative firm size measures (ln total assets): similar positive size premium; correlations: employees with total assets = 0.73; with sales = 0.82.
  - Results robust to inclusion/exclusion of pre-crisis TFP controls (Appendix Tables A6-A9).

### Policy-relevant implications and conclusions
- Main conclusions:
  - Smaller firms experienced a larger decline in post-crisis TFP growth vis-à-vis larger counterparts; effect progressively larger across medium, small, micro.
  - Impact disproportionately larger for firms in countries or with creditor banks facing more severe tightening in credit conditions (∆CDS, BLS evidence).
  - Access to stronger creditor banks (higher pre-crisis Tier 1 capitalization or CDS presence) mitigates the firm size premium.
  - Smaller firms reduced intangible investment and patenting more than larger firms following the crisis, linking credit constraints to TFP declines.
- Policy-relevant implications:
  - Improving creditor market access may help mitigate negative effects of adverse financial conditions on small-firm productivity.
  - Strengthening bank capitalization and supporting functioning of market-based creditor indicators (such as CDS presence) may reduce propagation of credit supply shocks to SMEs.
  - Policies targeting the extensive margin—facilitating access of small firms to stronger banks—could be particularly relevant to protect SME productivity in crises.
  - Tightening of credit market conditions during the crisis, coupled with limited access for smaller firms, may have contributed to persistent aggregate TFP decline and divergence between SMEs and large firms.

*Source: wpiea2020294-print-pdf - References, excerpt of the paper’s Introduction, data description, findings, and literature discussion.*

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

### wpiea2020294-print-pdf - References .............................................................................................................

### Introduction and research questions
- Investigates whether the post-Global Financial Crisis (GFC) productivity slowdown impacted firms unevenly across the size distribution and why.
- Two main questions:
  - Did smaller firms experience a larger decline in post-crisis TFP growth vis-à-vis their larger counterparts?
  - Can any observed productivity slowdown for smaller firms be explained by ex-ante limited credit market access?
- Defines micro, small, and medium-size enterprises following Eurostat as firms with employees no more than 10, 50, and 250, respectively; large firms are those with 250 or more employees.
- Final sample covers 8 European countries in the period before and after the GFC.
- Key event used to identify credit supply tightening: collapse of Lehman Brothers on September 15, 2008.

### Data and measurement
- Constructed a cross-country firm-level dataset by merging Orbis, AMADEUS, and Fitch Connect (Bankscope).
- Dataset contains:
  - Detailed firm-level input, output, and balance sheet information.
  - Information on each firm’s main bank and bank balance sheet information.
  - For a subset of firms, annual patent applications from PATSTAT.
- Measurement choices and features:
  - TFP estimated from detailed firm-level input and output data.
  - Credit supply tightening measured using:
    - Average CDS spread of banks in the firm’s home country around September 15, 2008.
    - Average CDS spread of the firm’s main creditor banks around the same period.
  - Complementary measures include bank lending survey–based measures of credit supply tightening.

### Main empirical findings
- Firm size premium of post-crisis TFP growth:
  - Smaller firms were more adversely affected than larger firms, with the impact progressively larger for medium, small, and micro firms relative to large firms.
  - The post-crisis decline in within-firm TFP growth was significantly larger among smaller firms than among larger counterparts.
  - The firm size premium is estimated within narrowly defined industry×country groups and persists after controlling for industry×country shocks.
  - The firm size premium holds after controlling for firm-level indicators of financial vulnerability and life cycle characteristics.
- Interaction with tightened credit conditions:
  - Smaller firms that faced more severe tightening of credit conditions during the crisis experienced larger TFP slowdowns.
  - Results are robust to alternative measures of credit supply tightening, including bank lending surveys.
- Firm size versus balance sheet vulnerabilities:
  - Firm size is an important factor for TFP vulnerability distinct from balance sheet characteristics.
  - Firm size effects strengthen when controlling for balance sheet characteristics.
  - Effects of balance sheet characteristics weaken when controlling for firm size.
  - Interpretation: firm size may be a more important and robust vulnerability indicator than common balance sheet measures.
- Extensive margin of credit market access:
  - Smaller firms were less likely to have relationships with better capitalized banks and banks with CDS presence.

### Credit market access channel and creditor strength
- Hypothesis: the firm size premium of post-crisis TFP growth is mitigated for firms with stronger credit market access.
- Two measures of creditor strength examined:
  - Bank capitalization: less capitalized banks face higher pressure to deleverage and cut lending when asset quality deteriorates.
  - Bank presence in the CDS market: CDS presence associated with sounder fundamentals, higher financing capacity, and lower borrowing cost.
- Findings:
  - The firm size premium of post-crisis TFP growth was mitigated by access to better capitalized banks and banks with CDS presence.
  - Evidence supports the role of creditor strength in cushioning firms—particularly small firms—from the adverse effects of tightened credit supply.

### Causality, timing, and robustness checks
- Causal interpretation arguments:
  - Firm size is highly persistent and predates the GFC; firm size is therefore an ex ante vulnerability unlikely to be correlated with unobserved firm characteristics conditional on controls.
  - Controls include industry×country fixed effects, firm-level balance sheet vulnerability indicators, firm age, and pre-crisis level of TFP.
  - In the data, firm size (measured by the log number of employees) has an autocorrelation of 0.95.
  - Confirmation that smaller firms did not on average experience slower TFP growth than larger firms before the crisis.
- Placebo test:
  - No evidence that the credit access channel operated in the post-2000 recession:
    - TFP growth of smaller firms did not decline more than larger counterparts after the 2000 recession.
    - Post-2000 results did not depend on the strength of creditor banks.
  - Interpretation: small firms’ credit market vulnerability is specific to the post-GFC episode in this study.

### Relationship to existing literature and contribution
- Bridges two strands of literature:
  - Studies using publicly listed firms that link firm size to financial position but underrepresent private and very small firms.
  - Studies using company registries or surveys covering the full firm distribution but lacking joint measures of finance and productivity.
- Complements prior work on financial frictions and productivity by focusing on within-firm TFP growth rather than only resource allocation across firms or R&D investment.
- Related empirical evidence:
  - Small firms are informationally opaque, reliant on bank financing, and more likely to be marginal borrowers relying on small local banks.
  - Prior studies cited include Gertler and Hubbard (1988), Custodio et al. (2013), Stein (2002), Rajan (1992), Aghion et al., Chodorow-Reich (2013), Siemer (2019), Huber (2018), Duval et al. (2019).

### Policy-relevant implications
- Improving creditor market access may help mitigate the negative effect of adverse financial conditions on small-firm productivity.
- Strengthening bank capitalization and supporting the functioning of market-based creditor indicators (such as CDS presence) may reduce propagation of credit supply shocks to SMEs.
- Policies targeting the extensive margin—facilitating access of small firms to stronger banks—could be particularly relevant to protect SME productivity in crises.

*Source: wpiea2020294-print-pdf - References, excerpt of the paper’s Introduction, data description, findings, and literature discussion.*

### Section III presents the empirical framework and results on small firm vulnerability. Section

### wpiea2020294-print-pdf - Section III presents the empirical framework and results on small firm vulnerability. Section

### II. DATA AND MEASUREMENT — Data sources and sample
- Firm-level data: Orbis historical database (Bureau van Dijk) merged from historical vintages.
- Bank-firm relationship: BANKER variable in the AMADEUS dataset listing up to five most important creditor banks per firm; BANKER sourced originally from Kompass.
- Creditor bank balance sheets matched to firms via Fitch Connect (formerly Bankscope) using probabilistic record linkage (Stata’s reclink2); matched 74 percent of all bank names.
- Final sample: 8 countries — Denmark, France, Germany, Netherlands, Portugal, Spain, Sweden, and the United Kingdom; selection criteria:
  - No significant change in firm coverage over sample period compared to Eurostat SBS (change in annual coverage does not exceed 15 percent; exception: Portugal increase between 2005-2006 led to excluding Portuguese firms in 2005).
  - Availability of BANKER variable.
- Sample coverage relative to Eurostat SBS:
  - France, Portugal, and Spain: gross output and employment coverage on average above 60 percent.
  - Denmark, Germany, Sweden: between 30 and 50 percent.
  - Netherlands: between 15 and 20 percent (stable).
- Data cleaning and transformations (following Gal 2013):
  - Nominal variables in U.S. dollars converted to local currency, deflated using OECD STAN local currency deflators, then converted to 2005 U.S. dollars using country-industry PPP from Inklaar et al (2005).
  - Financial firms and government-owned firms dropped; keep other sectors.
  - Keep firms that continuously exist over sample period (4 years before and after the GFC) — excludes entrants/exits to avoid attrition bias; note this restriction limits inference on aggregate TFP effects due to entry/exit.

### II. DATA AND MEASUREMENT — Measuring productivity
- Production function estimated using Wooldridge (2009) one-step efficient GMM control function approach.
- Estimated production function (as in source):
  - y_ijct = a_ijct + β_kj × k_ijct + β_lj × l_ijct−1 + ε_ijct
  - y, k, l are natural logs of value-added output, physical capital stock, and number of employees, respectively.
  - a is TFP in natural logarithm; output and physical capital stock expressed in real terms using country-industry price deflators.
- Firm-level TFP calculated as:
  - a_ijct = y_ijct − β̂_ks × k_ijct − β̂_ls × l_ijct−1
- TFP growth at time t computed as a_ijct − a_ijct−1 (log difference).

### III. SMALL FIRM VULNERABILITY AND PRODUCTIVITY GROWTH — Empirical framework
- Baseline differences-in-differences specification comparing pre- (2004-2007) and post-crisis (2008-2011) average TFP growth:
  - ∆TFPgrowth_i jc = α_jc + β Size_i + γ X_i + ε_ijc
  - Main size measures: (1) ln(number of employees), (2) dummies for micro (≤10 employees), small (≤50), medium (≤250) relative to large (≥250) per Eurostat.
- Controls X_i include:
  - Firm age (linear and quadratic), debt maturity (current liabilities to sales), leverage (total liabilities / total assets), liquidity (cash / total assets), earnings (ln EBITDA), pre-crisis TFP level.
- Fixed effects: four-digit-industry × country. Standard errors clustered at industry×country.
- Identification assumption: conditional on controls, remaining post-GFC TFP variation does not vary systematically with pre-crisis firm size within industry×country.
- Exploiting cross-country variation in credit supply shock:
  - Extended specification adds interaction with ∆CDS_c: the change in average CDS spread of domestic banks between September 7 and September 28, 2008 (7 days before and after Lehman collapse).
  - ∆CDS measured in basis points from 5-year CDS spreads (Markit); standardized to mean zero and unit standard deviation for interpretation.
  - Also replace country-level ∆CDS with average change in CDS of firm’s main creditor banks to exploit bank-firm relationship variation.

### III. SMALL FIRM VULNERABILITY AND PRODUCTIVITY GROWTH — Results on firm size premium (key findings)
- Descriptive statistics (Table 1 highlights):
  - Average pre-crisis TFP growth: 3 percent.
  - Post-crisis TFP growth dropped by 5 percentage points to -2 percent (∆TFP growth mean = -0.05).
  - Mean employees = 26; median = 7. Mean age = 19 years.
- Benchmark results (specification (2), Table 2):
  - ln(Employees) coefficient positive and significant: smaller firms experienced larger drop in post-crisis TFP growth relative to larger firms.
  - Column 3: a firm’s post-crisis TFP growth is on average 2.4 percentage points lower than its counterpart with about 170 percent more employees (one-unit increase in ln employees).
  - A one standard deviation decrease in employment (1.27 units of ln employees) accounts for 18 percent of a standard deviation in the drop in post-crisis TFP growth.
  - Micro, small, and medium dummies significantly negative and progressively more negative: micro firms experienced the largest TFP drop. Column 6 point estimates for declines relative to large firms:
    - Medium-sized: 1.4 percentage points lower.
    - Small: 3,1 percentage points lower.
    - Micro: 5.6 percentage points lower.
- Extended specification with country-level ∆CDS (Table 3):
  - ln(Employees) × ∆CDS interaction significantly positive: smaller firms’ TFP drop larger in countries where bank CDS spreads rose more.
  - Column 3: in a country with average increase in bank CDS spread (21 basis points), a one-unit decrease in ln employees associated with 0.9 percentage point drop in post-crisis TFP.
  - In a country with ∆CDS one-standard deviation higher than average (7.9 basis points higher), corresponding decline for one-unit decrease in ln employees is 2.4 percentage points.
  - Interaction estimates for micro/small/medium with ∆CDS significantly negative and progressively larger in magnitude: for a country with ∆CDS one-standard deviation above average, average decline relative to large firms:
    - Medium-sized: 0.9 percentage points lower.
    - Small: 2.1 percentage points lower.
    - Micro: 2.8 percentage points lower.
- Firm-level creditor-bank ∆CDS (Table 4):
  - Using firm creditor-bank CDS exposures yields qualitatively similar results though with smaller sample.
  - Column 3: a one-unit decrease in ln employees associated with 1.3 percentage point drop for firm whose creditor banks had average exposure to Lehman.
  - If creditor banks experienced ∆CDS one standard deviation higher than average, corresponding decline = 3.2 percentage points.
  - Column 6: with same creditor-bank ∆CDS increase, average decline in post-crisis TFP for medium, small, micro relative to large firms = 0.3, 1.9, and 6.1 percentage points respectively.

### IV. CHANNELS FOR SMALL FIRM VULNERABILITY — Credit market access of small firms (evidence)
- Hypothesis: smaller firms more vulnerable because of limited access to strong creditor banks; better-capitalized creditor banks mitigate firm size premium.
- Bank capitalization channel (specification (4), Table 5):
  - BankCapital dummy = 1 if average pre-crisis regulatory Tier 1 capital (percent of RWA) of firm’s credit banks is above median.
  - ln(Employees) × Bank capital interaction significantly negative: firm size premium smaller for firms with better-capitalized creditor banks.
  - Column 3: among firms with creditor banks capitalization above median, a one-unit decrease in ln employees associated with 1-percentage-point drop in post-crisis TFP; for below-median creditor banks, corresponding decline = 2.7 percentage points.
  - Effects strongest for small firms; micro and medium positive point estimates but not always statistically significant.
- CDS presence channel (specification (5), Table 6):
  - CDSpresence dummy = 1 if at least one creditor bank traded in single-name CDS.
  - ln(Employees) × CDS presence interaction significantly negative: access to banks with CDS presence reduces firm size premium.
  - Column 3: among firms with creditor banks with CDS presence, one-unit decrease in ln employees associated with 3.3-percentage-point drop; among firms without CDS presence, corresponding decline = 2.6 percentage points.
  - Column 6: access to banks with CDS presence reduces post-crisis TFP decline for medium, small, micro from 1.9 to 1.3, 4.2 to 2.6, and 7 to 5 percentage points respectively.
- Credit market access likelihood (equation (6), Tables 7 and 8):
  - Larger firms more likely to have access to high-capitalization creditor banks and banks with CDS presence.
  - Table 7 column 3: one-unit increase in ln employees associated with 29 percentage points increase in likelihood of access to high-capitalized bank (explains 110 percent of standard deviation in likelihood).
  - Table 7 column 6: micro firms 84 percent lower likelihood of access to better-capitalized banks; small and medium negative but less significant.
  - Table 8 column 3: one-unit increase in ln employees associated with 7 percentage points increase in likelihood of CDS presence (explains 16 percent of standard deviation).
  - Table 8 column 6: micro, small, medium likelihoods of CDS presence are 49, 45, and 19 percentage points lower respectively.

### IV. CHANNELS FOR SMALL FIRM VULNERABILITY — Intangible capital
- Hypothesis: smaller firms reduce intangible investment more when credit tightens, contributing to larger post-crisis TFP decline.
- Measures:
  - Intangible investment rate = change in stock of intangible assets / value added.
  - Intangible asset share = stock of intangible assets / (physical + intangible assets).
- Results (Tables 9 and 10):
  - Smaller firms cut intangible investment and reduce intangible asset share more than larger firms.
  - One-unit increase in ln employees associated with firm’s post-crisis intangible investment rate on average 0.1 percentage points lower and intangible share 0.4 percentage points lower than counterpart with about 170 percent more employees (one-unit ln employees).
  - Intangible investment rate reductions for medium, small, micro relative to large: 0.3, 0.7, and 0.8 percentage points lower respectively.
- Patent applications (Table 11):
  - Smaller firms reduce patent applications more than larger firms.
  - Column 3: one-unit decrease in ln employees associated with reduction of one patent application in post-crisis period.
  - Column 6: micro, small, medium firms reduce patent applications per year by 2.2, 1.9, and 1.8 respectively relative to large firms (sample mean patent applications ≈ 2 per year).

### V. DISCUSSION AND ROBUSTNESS — Placebo tests and size vs balance sheet vulnerabilities
- Placebo test (Table 12): using the 2000 dot-com bubble recession (no negative credit supply shock) — no association between firm size, creditor strength, and post-recession TFP growth differential.
- Horserace of explanatory factors (Tables 13 and 14):
  - Firm size effect strengthens when controlling for balance sheet characteristics; balance sheet effects weaken when controlling for firm size.
  - Example: one-unit decrease in ln employees associated with higher post-crisis TFP growth of 0.8 percentage point without controls and 2.4 percentage points with controls (Table 13).
  - Leverage effect diminishes by 32 percent when adding firm size; debt maturity and liquidity effects drop about 90 percent and lose significance when controlling for firm size.
  - Similar patterns hold in extended specifications with ∆CDS interactions.
  - Conclusion: firm size is an important and distinct factor for TFP vulnerability, partially explaining unconditional weaker-balance-sheet vulnerability.
- Robustness checks:
  - Top-bank instead of averaged creditor banks: similar results (Appendix).
  - Alternative credit condition measure: euro area Bank Lending Survey (BLS) — interaction of employment and credit tightening significantly positive (Appendix Table A3).
    - Column 3 (BLS): in a country with average tightening (72 percent of banks tightened), one-unit decrease in ln employees associated with 2.5 percentage point drop in post-crisis TFP; if tightening one-standard deviation higher (84 percent tightened), corresponding decline = 3.0 percentage points.
    - Micro firms exception: effect not statistically significant in one BLS estimate.
  - Alternative firm size: total assets (ln) — similar positive size premium (Appendix Tables A4-A5). Correlations: employees with total assets = 0.73; with sales = 0.82.
  - Robustness to controlling (or not) for pre-crisis TFP level: baseline and extended specifications not sensitive (Appendix Tables A6-A9).

### VI. CONCLUSION (implications)
- Main findings:
  - Smaller firms experienced a larger decline in post-crisis TFP growth vis-à-vis larger counterparts; effect progressively larger across medium, small, micro.
  - Impact disproportionately larger for firms in countries or with creditor banks facing more severe tightening in credit conditions (∆CDS, BLS evidence).
  - Access to stronger creditor banks (higher pre-crisis Tier 1 capitalization or CDS presence) mitigates the firm size premium.
  - Smaller firms reduce intangible investment and patenting more than larger firms following the crisis, linking credit constraints to TFP declines.
- Policy-relevant implication extracted from analysis:
  - Tightening of credit market conditions during the crisis, coupled with limited credit market access for smaller firms, may have contributed to large and persistent drop in aggregate TFP and the post-crisis TFP divergence between SMEs and large firms.
  - Mitigating factors include ensuring creditor bank strength and preserving SME access to well-capitalized lenders to limit long-run productivity losses via reduced intangible investment and innovation.

*Source: wpiea2020294-print-pdf (IMF working paper content provided in the input).*

### References

### References

### Cited works
- Ackerberg,  Daniel  A., Kevin Caves,  and  Garth Frazer.  2015.  Identification Properties of Recent Production Function Estimators. Econometrica 83:2411–51.
- Afonso, Gara, Anna Kovner, and Antoinette Schoar. 2011. Stressed, Not Frozen: The Federal Funds Market in The Financial Crisis. Journal of Finance 66:1109–39.
- Aghion,  Philippe,  George-Marios Angeletos,  Abhijit Banerjee,  and  Kalina Manova.  2010. Volatility  and Growth:  Credit Constraints  and the Composition  of Investment. Journal  of Monetary Economics 57:246–65.
- Aghion,  Philippe,  Philippe Askenazy,  Nicolas Berman,  Gilbert Cette,  and  Laurent Eymard. 2012.  Credit  constraints  and  the cyclicality  of R&D  investment:  Evidence  from  France. Journal of the European Economic Association 10:1001–1024.
- Alcalá, Francisco and Antonio Ciccone. 2004. Trade and Productivity. Quarterly Journal of Economics 119:613–46.
- Andrews, Dan, Chiara Criscuolo, and Peter N. Gal. 2016. The Best versus the Rest: The Global Productivity Slowdown,  Divergence  across  Firms  and  the  Role  of  Public  Policy. Technical Report, The Organisation for Economic Co-operation and Development.
- Augustin,  Patric, Marti  G.  Subrahmanyam, Dragon  Yongjun  Tang,  and Sarah  Qian  Wang.

*Source: wpiea2020294-print-pdf - References*

### 2014. Credit Default Swaps: A Survey. Foundations and Trends in Finance 9(1–2):1–196.

### 2014. Credit Default Swaps: A Survey. Foundations and Trends in Finance 9(1–2):1–196.

### Dependent variable and overall empirical setup
- Dependent variable: the difference in the average TFP growth between the pre- and post-crisis periods.
- All independent variables are measured at the pre-crisis period unless otherwise noted.
- Standard errors (in brackets) are clustered at the four digit industry*country level.
- Significance indicators: ***, **, and * indicate statistical significance at the 1, 5, and 10 percent respectively.
- Industry*country fixed effects (FEs) are included in all reported specifications.

### Robustness to bank capitalization (Table A1)
- Key coefficients (selected):
  - ln(Employees): 0.0107***; 0.0287***; 0.0289*** [0.000779] [0.00130] [0.00142]
  - ln(Employees) * Bank capital: -0.00542*; -0.0224***; -0.0227*** [0.00236] [0.00296] [0.00334]
  - Micro: -0.0294***; -0.0701***; -0.0677*** [0.00647] [0.00725] [0.00700]
  - Micro * Bank capital: 0.00238; 0.0101; 0.00807 [0.00908] [0.0156] [0.0152]
  - Debt maturity: -0.00423*; -0.00434*; -0.00605**; -0.00625** [0.00164] [0.00162] [0.00199] [0.00199]
  - Leverage: 0.0535***; 0.0517***; 0.0585***; 0.0605*** [0.00722] [0.00759] [0.00843] [0.00850]
  - Leverage * Bank capital: -0.0468**; -0.0454**; -0.0509***; -0.0531** [0.0133] [0.0148] [0.0143] [0.0157]
  - Liquidity: -0.0455**; -0.0499**; -0.0846***; -0.0784*** [0.0149] [0.0164] [0.0167] [0.0180]
  - TFP growth: -1.201***; -1.198***; -1.199***; -1.202***; -0.0701***; -0.0677*** [0.0144] [0.0129] [0.0133] [0.0146] [0.00725] [0.00700]
- Observations and fit:
  - Observations: 10,6017,997; 7,997; 10,7868,080; 8,080 (as reported)
  - R-squared: 0.620; 0.619; 0.619; 0.620; 0.613; 0.613

### Robustness to bank’s CDS presence (Table A2)
- Key coefficients (selected):
  - ln(Employees): 0.00841***; 0.0251***; 0.0262*** [0.000577] [0.00118] [0.00120]
  - ln(Employees) * CDS presence: -0.00294***; -0.00798***; -0.00787*** [0.000770] [0.00154] [0.00161]
  - Micro: -0.0240***; -0.0675***; -0.0689*** [0.00247] [0.00424] [0.00429]
  - Micro * CDS presence: 0.00762*; 0.0205***; 0.0194*** [0.00319] [0.00518] [0.00524]
  - Debt maturity: 0.00111; 0.00142; -0.00180*; -0.00170 [0.000901] [0.000907] [0.000890] [0.000898]
  - Leverage: 0.0417***; 0.0357***; 0.0458***; 0.0425*** [0.00257] [0.00236] [0.00249] [0.00246]
  - Liquidity: 0.0120; 0.00140; -0.0313**; -0.0379** [0.00798] [0.00893] [0.0104] [0.0116]
  - TFP growth: -1.205***; -1.204***; -1.205***; -1.203***; -1.205***; -1.206*** [0.0142] [0.0132] [0.0131] [0.0147] [0.0145] [0.0144]
- CDS presence: dummy equals 1 if the firm’s top creditor banks is traded in the CDS market.
- Observations and fit:
  - Observations: 207,734; 172,853; 172,853; 211,636; 174,610; 174,610
  - R-squared: 0.621; 0.632; 0.632; 0.620; 0.626; 0.626

### Robustness to credit condition measures (Table A3)
- Credit tightening defined as the change in credit standards in 2008Q4 from euro area bank lending survey.
- Key coefficients (selected):
  - ln(Employees): 0.00312***; 0.0140***; 0.0153*** [0.000586] [0.00176] [0.00187]
  - ln(Employees) * Credit tightening: 0.0158***; 0.0350***; 0.0353*** [0.00268] [0.00983] [0.0104]
  - Small * Credit tightening: -0.0681***; -0.168***; -0.167*** [0.0101] [0.0258] [0.0265]
  - Medium * Credit tightening: -0.0377***; -0.0728***; -0.0723*** [0.00985] [0.0144] [0.0147]
  - Leverage * Credit tightening: 0.0705***; 0.0665***; 0.0542***; 0.0546*** [0.0152] [0.0140] [0.0146] [0.0143]
  - Liquidity * Credit tightening: -0.407**; -0.442***; -0.371***; -0.381*** [0.125] [0.118] [0.0920] [0.0906]
  - TFP growth: -1.197***; -1.191***; -1.193***; -1.194***; -1.193***; -1.194*** [0.00522] [0.00511] [0.00508] [0.00548] [0.00561] [0.00558]
- Observations and fit:
  - Observations: 363,308; 286,133; 286,133; 368,643; 288,788; 288,788
  - R-squared: 0.606; 0.615; 0.616; 0.604; 0.608; 0.608

### Robustness to firm size (Table A4)
- Key coefficients (selected):
  - ln(Total assets): -0.00394***; -0.00577***; 0.0127***; 0.0118*** [0.000571] [0.000662] [0.00266] [0.00269]
  - Age: 0.00143***; 0.000656*** [0.000113] [0.000121]
  - Age squared: -1.33e-05***; -7.03e-06*** [1.38e-06] [1.41e-06]
  - Debt maturity: -0.00959***; -0.00946*** [0.00162] [0.00162]
  - Leverage: -0.0267***; -0.0234*** [0.00311] [0.00303]
  - Liquidity: -0.212***; -0.208*** [0.0350] [0.0346]
- Observations and fit:
  - Observations: 452,538; 452,538; 359,928; 359,928
  - R-squared: 0.056; 0.058; 0.082; 0.082

### Robustness to firm size and exposure to the collapse of Lehman Brothers (Table A5)
- ∆CDS is the standardized change in the country-level CDS between the weeks before and after the Lehman bankruptcy (average of changes in CDS spread of all domestic banks).
- Key coefficients (selected):
  - ln(Total assets): -0.00271***; -0.00361***; 0.00347; 0.00305 [0.000620] [0.000651] [0.00198] [0.00199]
  - ln(Total assets) * ∆CDS: -0.000379; -0.00147*; 0.00546**; 0.00500* [0.000667] [0.000732] [0.00190] [0.00194]
  - Age: 0.000848***; 0.000394** [0.000140] [0.000150]
  - Age * ∆CDS: 0.000647***; 0.000158 [0.000131] [0.000137]
  - Debt maturity * ∆CDS: -0.00686**; -0.00682** [0.00229] [0.00229]
  - Leverage * ∆CDS: -0.0208***; -0.0190*** [0.00365] [0.00362]
  - Liquidity * ∆CDS: -0.172***; -0.169*** [0.0200] [0.0197]
- Observations and fit:
  - Observations: 209,294; 209,294; 174,781; 174,781
  - R-squared: 0.060; 0.062; 0.089; 0.089

### Robustness to pre-crisis TFP control (baseline specification) (Table A6)
- Key coefficients (selected):
  - ln(Employees): 0.00630***; 0.0290***; 0.0291*** [0.000654] [0.00172] [0.00173]
  - Micro firms: -0.0331***; -0.0989***; -0.0989*** [0.00237] [0.00595] [0.00586]
  - Small firms: -0.0284***; -0.0694***; -0.0700*** [0.00232] [0.00474] [0.00470]
  - Medium-sized firms: -0.0132***; -0.0338***; -0.0348*** [0.00211] [0.00311] [0.00312]
  - Debt maturity: -0.00117; -0.00111; -0.00473***; -0.00478*** [0.000886] [0.000901] [0.000952] [0.000952]
  - Leverage: -0.0407***; -0.0405***; -0.0333***; -0.0310*** [0.00329] [0.00328] [0.00317] [0.00315]
  - Liquidity: -0.187***; -0.187***; -0.235***; -0.231*** [0.0222] [0.0224] [0.0237] [0.0238]
- Observations and fit:
  - Observations: 452,644; 358,530; 358,530; 458,940; 361,614; 361,614
  - R-squared: 0.056; 0.095; 0.095; 0.056; 0.087; 0.087

### Robustness to pre-crisis TFP control: exposure to Lehman (Table A7)
- Key coefficients (selected):
  - ln(Employees): 0.000521; 0.00936***; 0.00931*** [0.000891] [0.00163] [0.00165]
  - ln(Employees) * ∆CDS: 0.00805***; 0.0197***; 0.0201*** [0.000737] [0.00167] [0.00172]
  - Micro: -0.0105**; -0.0422***; -0.0419*** [0.00326] [0.00545] [0.00541]
  - Micro * ∆CDS: -0.0271***; -0.0593***; -0.0596*** [0.00328] [0.00570] [0.00570]
  - Debt maturity: -0.0239***; -0.0207***; -0.00109; -0.00109 [0.00310] [0.00409] [0.00264] [0.00264]
  - Leverage: -0.158***; -0.141***; -0.0201***; -0.0189*** [0.0114] [0.0161] [0.00400] [0.00408]
  - Liquidity pre-crisis: 4.04e-07***; 4.04e-07*; -0.163***; -0.162*** [1.07e-07] [1.89e-07] [0.0176] [0.0177]
  - Liquidity pre-crisis * ∆CDS: 3.36e-07***; 3.50e-07*; -0.174***; -0.173*** [8.95e-08] [1.59e-07] [0.0165] [0.0165]
- Observations and fit:
  - Observations: 208,681; 173,727; 173,727; 212,577; 175,487; 175,487
  - R-squared: 0.063; 0.109; 0.109; 0.062; 0.100; 0.100

### Robustness to pre-crisis TFP control: bank capitalization (Table A8)
- Key coefficients (selected):
  - ln(Employees): 0.0176***; 0.0451***; 0.0449*** [0.00131] [0.00356] [0.00363]
  - ln(Employees) * Bank capital: -0.00903**; -0.0252***; -0.0251*** [0.00299] [0.00732] [0.00729]
  - Micro: -0.0690***; -0.142***; -0.137*** [0.00572] [0.0102] [0.0104]
  - Micro * Bank capital: 0.0299*; 0.0354; 0.0344 [0.0149] [0.0328] [0.0314]
  - Debt maturity: -0.00417; -0.00419; -0.00715**; -0.00735** [0.00240] [0.00242] [0.00266] [0.00268]
  - Leverage: -0.0592***; -0.0579***; -0.0554***; -0.0487*** [0.00814] [0.00872] [0.00809] [0.00867]
  - Liquidity: -0.482***; -0.479***; -0.541***; -0.522*** [0.0534] [0.0542] [0.0481] [0.0495]
  - Liquidity * Bank capital: 0.280**; 0.270**; 0.312***; 0.287** [0.0939] [0.0958] [0.0901] [0.0932]
  - Bank capital: 0.0415**; -0.162*; -0.167**; -0.00397; -0.212**; -0.211** [0.0156] [0.0661] [0.0631] [0.0117] [0.0814] [0.0783]
- Observations and fit:
  - Observations: 24,458; 20,542; 20,542; 25,304; 20,940; 20,940
  - R-squared: 0.079; 0.128; 0.128; 0.077; 0.116; 0.116

### Robustness to pre-crisis TFP control: CDS presence (Table A9)
- CDS presence: dummy equals 1 if at least one of the firm’s creditor banks is traded in the CDS market.
- Key coefficients (selected):
  - ln(Employees): 0.00836***; 0.0336***; 0.0336*** [0.000831] [0.00215] [0.00216]
  - ln(Employees) * CDS presence: -0.00206**; -0.0104***; -0.0102*** [0.000726] [0.00217] [0.00213]
  - Micro: -0.0372***; -0.117***; -0.115*** [0.00448] [0.00849] [0.00818]
  - Micro * CDS presence: 0.00607; 0.0325***; 0.0317*** [0.00471] [0.00936] [0.00896]
  - Debt maturity: -0.000501; -0.000484; -0.00415*; -0.00425* [0.00167] [0.00165] [0.00176] [0.00175]
  - Leverage: -0.0481***; -0.0476***; -0.0437***; -0.0402*** [0.00459] [0.00476] [0.00472] [0.00500]
  - Leverage * CDS presence: 0.0162*; 0.0155*; 0.0162*; 0.0141 [0.00682] [0.00716] [0.00720] [0.00771]
  - Liquidity: -0.202***; -0.201***; -0.251***; -0.244*** [0.0442] [0.0452] [0.0500] [0.0503]
  - Liquidity * CDS presence: 0.0709; 0.0695; 0.0927; 0.0869 [0.0632] [0.0642] [0.0729] [0.0732]
  - CDS presence: 0.00265; -0.100***; -0.100***; -0.00517; -0.125***; -0.122*** [0.00219] [0.0145] [0.0142] [0.00442] [0.0218] [0.0208]
- Observations and fit:
  - Observations: 208,681; 173,727; 173,727; 212,577; 175,487; 175,487
  - R-squared: 0.062; 0.101; 0.101; 0.061; 0.092; 0.093

*Source: 2014. Credit Default Swaps: A Survey. Foundations and Trends in Finance 9(1–2):1–196. (Appendix tables and notes as provided in the source PDF.)*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020294-print-pdf.pdf_