## _wp1498

## Source details

**Canonical URL:** [_wp1498](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp1498.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp1498.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp1498.pdf.json)

---

### Key findings on SMEs' role and recent performance
- SMEs account for about 99 percent of the total number of enterprises in the EU.
- SMEs' estimated share in the EU non-financial business sector:
  - 58 percent by value added.
  - 66 percent by employment.
- In 2012, the SMEs sector was:
  - 2 percent below its pre-crisis employment level (2008 = 100 baseline).
  - 10 percent below its pre-crisis value added level (2008 = 100 baseline).
- Cross-country evidence (2008–12):
  - Countries with higher prevalence of SMEs experienced, on average, slower real GDP growth in 2008–12.
  - At end-2012, real GDP had surpassed pre-crisis (2008) levels in only nine out of twenty seven EU members.

### Financial conditions and SME vulnerability
- Recent tightening in lending conditions observed:
  - Banks tightened credit supply and raised credit spreads.
  - Banks applied tighter lending requirements.
- SME-specific vulnerabilities:
  - Fewer assets eligible as loan collateral.
  - Less access to alternative financing (debt/equity issuance, international markets, support from parent companies).
- Consequence: Significant funding pressures for SMEs with adverse effects on investment plans and activity.

### Empirical approach and specification
- Objective: Examine extent to which tight financial conditions faced by SMEs affected pace of economic recovery (post-global financial crisis).
- Data and scope:
  - SMEs share in non-financial business sector’s value added used as prevalence indicator (average, 2002–2012).
  - Quarterly data 2002q1–2012q4.
- Key variables:
  - Growth = GDP growth (y-o-y) in quarter t in country i.
  - SMEs = countries’ SMEs share in the non-financial business sector’s value added (average, 2002–2012).
  - Dum_crisis = 1 for 2008q4–2009q4, zero otherwise.
  - Dum_recovery = 1 for 2010q1–2012q4, zero otherwise.
  - Controls: US_Growth, VIX, Pub_Debt, Corp_Debt.
- Methodologies: OLS, Fixed Effects, Dynamic Panel.
- Purpose of interaction terms: Assess impact of SMEs prevalence in crisis and recovery periods via Dum_crisis*SMEs and Dum_recovery*SMEs.

### Main econometric results (selected)
- Dum_recovery * SMEs coefficients (Specification A, Table 1):
  - -0.087* (OLS)
  - -0.051*** (Fixed Effects)
  - -0.034** (Dynamic Panel)
- Dum_crisis * SMEs: reported positive and/or significant in some specifications (example: 0.069*** in one column).
- Other notable coefficients:
  - GDP Growth(-1): 0.694*, 0.683* (reported in columns).
  - Pub_Debt: -0.042*, -0.040*, -0.071*, -0.070*, -0.028*, -0.029*.
  - US_Growth: 0.384*, 0.387*, 0.369*, 0.369*, 0.365*, 0.352*.
  - VIX: -0.035* (negative and significant in relevant columns).
  - Corp_Debt: -0.007*, -0.007*, -0.007**, -0.007** (negative and significant in some columns).
  - Constants reported (examples): 5.530*, 5.350*, 7.111*, 7.079*, 2.270*, 2.357*.
- Sample size and fit:
  - # obs.: 822, 822, 822, 822, 966, 966 (across columns).
  - R-squared (overall): 0.570, 0.578, 0.535, 0.543 (reported).
- Significance notation: * significant at 1 percent, ** significant at 5 percent, *** significant at 10 percent.
- Interpretation:
  - Dum_recovery * SMEs yields negative and significant coefficients across methodologies, indicating countries with higher SME prevalence experienced weaker growth in 2010–12 relative to the sample average.
  - Once interaction is included, Dum_recovery coefficient for the overall sample becomes positive and significant in two out of three specifications.

### Specification B (SMEs_High) and Table 2 findings
- Specification: interaction uses SMEs_High = 1 for countries with share of SMEs above the sample’s median (60 percent of the non-financial business sector’s value added) and 0 otherwise.
- Sample period: 2002q1–2012q4.
- Selected estimates (Table 2, dependent variable: GDP Growth, y-o-y):
  - GDP Growth(-1): 0.690*; 0.686*
  - Dum_crisis: -4.856*; -4.653*; -4.700*; -4.649*; -0.096; -0.118
  - Dum_recovery: -1.055*; -0.526**; -0.471***; -0.136; 0.160; 0.377**
  - Dum_crisis*SMEs_High: -0.481; -0.130; -0.009
  - Dum_recovery*SMEs_High: -1.277*; -0.849**; -0.447**
  - US_Growth: 0.384*; 0.387*; 0.369*; 0.372*; 0.350*; 0.352*
  - Pub_Debt: -0.042*; -0.041*; -0.071*; -0.069*; -0.028*; -0.028*
  - VIX: -0.035*; -0.035*
  - Corp_Debt: -0.007*; -0.007*; -0.007*; -0.007**
  - Constant: 5.530*; 5.372*; 7.111*; 7.011*; 2.317*; 2.317*
  - # obs.: 822; 822; 822; 822; 966; 966
  - R-squared (overall): 0.570; 0.580; 0.535; 0.546
- Inference: Estimations consistent with prior results and reaffirm that countries with high prevalence of SMEs registered, on average, negative economic growth in 2010–12.

### Credit prevalence and post-crisis credit growth correlation
- Negative correlation between cumulative growth of credit to non-financial corporates in 2008–12 and prevalence of SMEs.
- Scatter regressions (2008–12 cumulative growth):
  - Real credit: y = -0.2092x + 60.943, R² = 0.199
  - Nominal credit: y = -0.2314x + 62.698, R² = 0.3001
- Pattern: Countries with high (low) prevalence of SMEs registered, on average, lower (higher) credit growth (real and nominal) in 2008–12.
- Data note: BIS data used on credit to non-financial corporations (all sectors) since comparable SME credit stock series not available for all EU countries; possible valuation issues where foreign currency lending is extensive.

### Two-stage approach to measure SME credit supply shocks
- Objective: Estimate effects of credit supply shocks faced by SMEs on economic growth.
- First stage:
  - Use bank lending surveys to create measure for supply of credit to SMEs by purging reported lending standards from reported changes in loan demand by SMEs.
  - Surveys report net percentage of banks (weighted by market shares) tightening/loosening lending standards.
  - Caveats: surveys are qualitative and differ slightly across countries.
- First-stage data and sample:
  - Bank lending surveys from 15 central banks.
  - Estimation sample: 15 countries, 516 observations, period 2003q1–2012q4 (shorter samples for Slovenia, Malta, Cyprus, UK; Hungary mixed frequency adjustment described).
- First-stage dynamic panel (Arellano-Bond) result (z-statistics):
  - ΔS_{i,t} = 4.995 + 0.643*ΔS_{i,t-1} - 0.092*ΔD_{i,t} + u_{i,t}
    - Coefficients (t-statistics): (4.68) (10.44) (-2.20)
  - Interpretation: mutual factors affect both credit demand and lending standards; high persistence in lending standards (lagged coefficient 0.643).

### Panel VAR specification and endogenous variables
- Residuals from first-stage (purged lending standards) used as proxy for supply-driven constraints to SME lending.
- Panel VAR estimated using Love and Zicchino (2006) program, fixed effects with forward mean-differencing (Helmert), GMM instruments.
- Endogenous variables vector (five variables):
  - Residual from Eq. 2 (adjusted changes in lending standards for SMEs)
  - Real GDP growth (yoy)
  - Inflation (yoy change of GDP deflator)
  - Real growth of credit to non-financial corporates (yoy)
  - Interest rate spread: gap between interest rate on new small loans (less than 1 million euro, maturity up to one year) and free-risk rate (policy rate)
- Estimation details:
  - Sample period 2003q1–2012q4
  - Three lags of endogenous variables used
  - Ordering: adjusted lending standards first (credit supply shock placed first), then real GDP growth and inflation, then credit growth and interest spreads
  - Dynamic behavior assessed via impulse response functions with shocks orthogonalized using Cholesky decomposition
  - Confidence intervals via Monte Carlo simulations

### Impulse response results and macroeconomic effects
- Shock definition: one standard deviation to credit supply measure ≈ adjusted “net tightening” of about 18 percent.
- Impact on real GDP growth:
  - An adverse credit supply shock leads to a contraction lasting six quarters with a trough reaching ½ percentage points in the fourth quarter.
- Impact on real growth of credit to non-financial corporations:
  - Prolonged contraction, reaching a trough after eight quarters.
  - Slow adjustment reflects that credit growth is measured on the stock (maturity structure delays change in stock even when new lending is curtailed).
- Interest rate spread:
  - Shock leads to wider interest rate spreads, but shock captures non-price tightening (collateral, loan size, maturity, non-interest charges), implying many loan applications may be rejected even if spread increases are modest.
- Robustness checks (two alternative VARs):
  - Alternative I (use unadjusted ΔS): sharper contraction of real GDP, trough of 0.6 percentage point in the 3rd quarter; stronger responses of inflation, credit growth, and interest spread than baseline.
  - Alternative II (change VAR ordering to allow contemporaneous effects on adjusted lending standards): more moderate impact on real GDP, trough of 0.4 percentage points in the 4th quarter; weaker feedback effects on other variables.
  - Overall: similar patterns across baseline and alternatives though magnitudes differ.

### Sample split (High vs Low SMEs share) and sub-sample VAR results
- Split threshold: Median = 62 percent (SMEs share in non-financial business sector’s value added).
- Observations for Lithuania, the Netherlands, and Slovenia that crossed 62 percent over time were split between sub-samples.
- Table 3A composition (Annex):
  - High share: Cyprus, Greece, Italy, Spain, Portugal, Malta, Lithuania (2006q1–2012q4), the Netherlands (2006q1–2012q4), Slovenia (2007q1–2012q4).
  - Low share: Austria, Germany, France, Hungary, Lithuania (2003q1–2005q4), the Netherlands (2003q1–2005q4), Poland, Slovenia (2003q1–2006q4), UK.
- Key empirical findings:
  - An adverse credit supply shock (standardized to be the same in both sub-samples) has a significant impact on GDP growth in both sub-samples.
  - Impact is much more prominent in the “High share” sub-sample.
- Quantified medium-term cumulative impacts:
  - In the “High share” sub-sample: cumulative contraction of 2¾ percent in real output from a credit supply shock (net tightening by 18 percent).
  - In the “Low share” sub-sample: cumulative contraction of 1¼ percent in real output from the same shock.
- Suggested mechanism: stronger feedback effects from weak economic activity to banks’ balance sheets in High share countries may prolong tightening and keep interest spreads higher for longer.
- Corroborating patterns: pace of economic recovery and credit growth during 2010–2012 negatively correlated with prevalence of SMEs across EU countries.

### Caveats, robustness, and interpretation
- Data limitations:
  - SMEs shares based on actual data for 2002–2010 and nowcasts for 2011–12.
  - SMEs shares capture formal sector prevalence and may be biased downwards if SMEs are more prevalent in the informal economy.
  - Potential inconsistencies in SME definitions (EU definition vs. bank lending surveys unit/metric inconsistency noted).
- Bank lending survey limitations:
  - Qualitative nature does not capture magnitude of tightening/easing.
  - Differences in survey questions across countries may create inconsistencies.
- Identification and interpretation caveats:
  - First-stage purging may not fully solve endogeneity (demand may respond to past supply changes).
  - Relatively high impact on GDP growth may reflect sample composition (several crisis and near-crisis cases) and possible spillovers of SME-targeted tightening to other sectors.
  - “High share” group includes four countries at the epicenter of the crisis (Italy, Cyprus, Portugal, Spain), so impacts may be overstated if they capture broader financial intermediation problems, tighter fiscal policies, or negative confidence effects.

### Policy implications and recommendations
- Maintain adequate access to finance for SMEs to support sustainable economic recovery, particularly in countries with high prevalence of SMEs.
- Differentiate policy support between:
  - High-productivity and solvent SMEs; and
  - SMEs that lack a sustainable business plan.
- Short- to medium-term measures (limited success potential):
  - Cheap long-term liquidity provision by central banks (examples cited in text).
  - Expansion of state guarantee schemes, especially if conditional or linked to new lending to SMEs.
- Structural and financial intermediation measures:
  - Reduce fragmentation in credit markets and restore financial intermediation.
  - Strengthen banks’ balance sheets through a faster cleanup of bad assets and measures to ensure adequate capital and liquidity positions.
  - Increase availability of non-bank financing sources, including development of securitization markets for SMEs loans.

*Source: _wp1498 - References (PDF), sections I–II and Tables 1–3A, and Annex as provided.*

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

### _wp1498 - References

### Key findings on SMEs' role and recent performance
- SMEs account for about 99 percent of the total number of enterprises in the EU.
- SMEs' estimated share in the EU non-financial business sector:
  - 58 percent by value added.
  - 66 percent by employment.
- In 2012, the SMEs sector was:
  - 2 percent below its pre-crisis employment level (2008 = 100 baseline).
  - 10 percent below its pre-crisis value added level (2008 = 100 baseline).
- Cross-country evidence (2008–12):
  - Countries with higher prevalence of SMEs experienced, on average, slower real GDP growth in 2008–12.
  - At end-2012, real GDP had surpassed pre-crisis (2008) levels in only nine out of twenty seven EU members.

### Financial conditions and SME vulnerability
- Recent tightening in lending conditions observed:
  - Banks tightened credit supply and raised credit spreads (Figure 2).
  - Banks applied tighter lending requirements (Figure 3).
- SME-specific vulnerabilities highlighted:
  - Fewer assets eligible as loan collateral.
  - Less access to alternative financing (debt/equity issuance, international markets, support from parent companies).
- Consequence: Significant funding pressures for SMEs with adverse effects on investment plans and activity.

### Empirical approach and specification
- Objective: Examine extent to which tight financial conditions faced by SMEs affected pace of economic recovery (post-global financial crisis).
- Data and scope:
  - SMEs share in non-financial business sector’s value added used as prevalence indicator (average, 2002–2012).
  - Quarterly data 2002q1–2012q4.
- Key variables and constructs:
  - Growth = GDP growth (y-o-y) in quarter t in country i.
  - SMEs = countries’ SMEs share in the non-financial business sector’s value added (average, 2002–2012).
  - Dum_crisis = 1 for 2008q4–2009q4, zero otherwise.
  - Dum_recovery = 1 for 2010q1–2012q4, zero otherwise.
  - Controls include US_Growth, VIX, Pub_Debt, and Corp_Debt.
- Methodologies employed:
  - OLS, Fixed Effects, and Dynamic Panel estimations.
- Purpose of interaction terms:
  - Assess impact of SMEs prevalence in crisis and recovery periods via Dum_crisis*SMEs and Dum_recovery*SMEs.

### Main econometric results (selected)
- Table 1 (Prevalence of SMEs and Economic Recovery, 2002q1–2012q4, Specification A):
  - Dum_recovery * SMEs coefficients:
    - -0.087* (OLS)
    - -0.051*** (Fixed Effects)
    - -0.034** (Dynamic Panel)
  - Dum_crisis * SMEs: reported positive and/or significant in some specifications (e.g., 0.069*** in one column).
  - GDP Growth(-1): 0.694* and 0.683* reported in columns.
  - Pub_Debt coefficients: -0.042*, -0.040*, -0.071*, -0.070*, -0.028*, -0.029* across columns (significant with expected negative sign).
  - US_Growth reported positive and significant (e.g., 0.384*, 0.387*, 0.369*, 0.369*, 0.365*, 0.352*).
  - VIX reported negative and significant in relevant columns (e.g., -0.035*).
  - Corp_Debt reported negative and significant in some columns (e.g., -0.007*, -0.007*, -0.007**, -0.007**).
  - Constants reported (e.g., 5.530*, 5.350*, 7.111*, 7.079*, 2.270*, 2.357*).
  - # obs.: 822 822 822 822 966 966.
  - R-squared (overall): 0.570 0.578 0.535 0.543.
  - Significance notation: * significant at 1 percent, ** significant at 5 percent, *** significant at 10 percent.
- Interpretation:
  - Interaction Dum_recovery * SMEs yields negative and significant coefficients across methodologies, indicating countries with higher SME prevalence experienced weaker growth in 2010–12 relative to the sample average.
  - Once interaction is included, Dum_recovery coefficient for the overall sample becomes positive and significant in two out of three specifications.

### Panel VAR and shock analysis
- A panel VAR was estimated including a measure of banks’ credit supply (from bank lending surveys).
- Main insight:
  - An adverse credit supply shock faced by SMEs has a negative effect on economic activity.
  - The negative impact is larger in countries with a higher SMEs sector prevalence.
- Implication: Tight credit supply specifically targeted or disproportionately affecting SMEs slows macroeconomic recovery, with larger effects where SMEs compose a greater share of the non-financial business sector.

### Caveats and data limitations
- SMEs shares are based on actual data for 2002–2010 and nowcasts for 2011–12; lack of actual recent data poses challenges for assessing crisis implications on SMEs.
- Reported SMEs shares capture formal sector prevalence and may be biased downwards if SMEs are more prevalent in the informal economy.
- Potential inconsistencies in SME definitions:
  - EU definition: firms with less than 250 employees and an annual turnover of €50 million or a balance sheet total of €43 million.
  - Bank lending surveys: an enterprise is normally considered large if its annual net turnover is more than €50 (possible unit/metric inconsistency noted).

### Policy implications and conclusions
- Results underline the important contribution of SMEs to economic activity in Europe.
- Alleviating financing constraints for SMEs is highlighted as important to support a faster economic recovery.
- The evidence suggests targeted policies to ease SME access to credit could mitigate the larger negative macroeconomic impact in high-SME countries.

*Source: _wp1498 - References (PDF), sections I–II and Table 1 as provided.*

### 15.      For robustness, we also estimate an alternative specification that, instead of the

### _wp1498 - 15.      For robustness, we also estimate an alternative specification that, instead of the

### Specification B: SMEs_High interaction and Table 2 findings
- Specification: interaction term uses SMEs_High, a dummy = 1 for countries with share of SMEs above the sample’s median (60 percent of the non-financial business sector’s value added) and 0 otherwise.
- Sample period for Table 2: 2002q1–2012q4.
- Selected coefficient estimates (as presented in Table 2, dependent variable: GDP Growth, y-o-y):
  - GDP Growth(-1): 0.690*; 0.686*
  - Dum_crisis: -4.856*; -4.653*; -4.700*; -4.649*; -0.096; -0.118
  - Dum_recovery: -1.055*; -0.526**; -0.471***; -0.136; 0.160; 0.377**
  - Dum_crisis*SMEs_High: -0.481; -0.130; -0.009
  - Dum_recovery*SMEs_High: -1.277*; -0.849**; -0.447**
  - US_Growth: 0.384*; 0.387*; 0.369*; 0.372*; 0.350*; 0.352*
  - Pub_Debt: -0.042*; -0.041*; -0.071*; -0.069*; -0.028*; -0.028*
  - VIX: -0.035*; -0.035*
  - Corp_Debt: -0.007*; -0.007*; -0.007*; -0.007**
  - Constant: 5.530*; 5.372*; 7.111*; 7.011*; 2.317*; 2.317*
  - # obs.: 822; 822; 822; 822; 966; 966
  - R-squared (overall): 0.570; 0.580; 0.535; 0.546
- Significance legend from table: * significant at 1 percent, ** significant at 5 percent, *** significant at 10 percent.
- Key inference from Table 2: estimations are consistent with prior results and reaffirm that countries with high prevalence of SMEs registered, on average, negative economic growth in 2010–12.

### Credit prevalence and correlation with post-crisis credit growth
- Negative correlation documented between cumulative growth of credit to non-financial corporates in the post crisis period and the prevalence of SMEs.
- Scatter regressions (2008–12 cumulative growth):
  - Real credit: y = -0.2092x + 60.943, R² = 0.199
  - Nominal credit: y = -0.2314x + 62.698, R² = 0.3001
- Empirical pattern: countries with high (low) prevalence of SMEs registered, on average, lower (higher) credit growth (in both real and nominal terms) in 2008–12.
- Data note: BIS data used on credit to non-financial corporations (all sectors) since comparable SME credit stock series not available for all EU countries; possible valuation issues where foreign currency lending is extensive.

### Two-stage approach to measure SME credit supply shocks
- Objective: estimate effects of credit supply shocks faced by SMEs on economic growth.
- First stage:
  - Use bank lending surveys to create measure for supply of credit to SMEs by purging reported lending standards from reported changes in loan demand by SMEs.
  - Surveys report net percentage of banks (weighted by market shares) tightening/loosening lending standards.
  - Caveats: surveys are qualitative and differ slightly across countries.
- Data for first-stage regression:
  - Bank lending surveys from 15 central banks (countries listed in source).
  - Estimation sample: 15 countries, 516 observations, period 2003q1–2012q4 (shorter samples for Slovenia, Malta, Cyprus, UK; Hungary mixed frequency adjustment described).
- First-stage dynamic panel (Arellano-Bond) estimation result (z-statistics in parenthesis):
  - ΔS_{i,t} = 4.995 + 0.643*ΔS_{i,t-1} - 0.092*ΔD_{i,t} + u_{i,t}
    - Coefficients (t-statistics): (4.68) (10.44) (-2.20)
  - Interpretation: mutual factors affect both credit demand and lending standards; high persistence in lending standards (lagged coefficient 0.643).

### Panel VAR specification and endogenous variables
- Method:
  - Residuals from first-stage (purged lending standards) used as proxy for supply-driven constraints to SME lending.
  - Panel VAR estimated using Love and Zicchino (2006) program, fixed effects with forward mean-differencing (Helmert), GMM instruments.
- Endogenous variables vector (five variables):
  - Residual from Eq. 2 (adjusted changes in lending standards for SMEs)
  - Real GDP growth (yoy)
  - Inflation (yoy change of GDP deflator)
  - Real growth of credit to non-financial corporates (yoy)
  - Interest rate spread: gap between interest rate on new small loans (less than 1 million euro, maturity up to one year) and free-risk rate (policy rate)
- Estimation details:
  - Sample period 2003q1–2012q4
  - Three lags of endogenous variables used
  - Ordering: adjusted lending standards first (credit supply shock placed first), then real GDP growth and inflation, then credit growth and interest spreads
  - Dynamic behavior assessed via impulse response functions with shocks orthogonalized using Cholesky decomposition
  - Confidence intervals via Monte Carlo simulations

### Impulse response results and macroeconomic effects
- Shock definition: one standard deviation to credit supply measure ≈ adjusted “net tightening” of about 18 percent.
- Impact on real GDP growth:
  - An adverse credit supply shock leads to a contraction lasting six quarters with a trough reaching ½ percentage points in the fourth quarter.
- Impact on real growth of credit to non-financial corporations:
  - Prolonged contraction, reaching a trough after eight quarters.
  - Slow adjustment reflects that credit growth is measured on the stock (maturity structure delays change in stock even when new lending is curtailed).
- Interest rate spread:
  - Shock leads to wider interest rate spreads, but the shock also captures non-price tightening (collateral, loan size, maturity, non-interest charges), suggesting many loan applications may be rejected even if spread increases are modest.
- Robustness checks (two alternative VARs, estimated over 2003q1–2012q4 with three lags):
  - Alternative I: use unadjusted measure ΔS instead of purged residual.
    - Result: sharper contraction of real GDP, reaching a trough of 0.6 percentage point in the 3rd quarter; stronger responses of inflation, credit growth, and interest spread than baseline.
  - Alternative II: change VAR ordering to allow real GDP growth and inflation to have contemporaneous effects on adjusted lending standards.
    - Result: more moderate impact on real GDP, reaching a trough of 0.4 percentage points in the 4th quarter; weaker feedback effects on other variables.
  - Overall: patterns across baseline and alternatives are similar though magnitudes differ.

### Caveats and interpretation
- Bank lending survey limitations:
  - Qualitative nature does not capture magnitude of tightening/easing.
  - Differences in survey questions across countries may create inconsistencies.
- First-stage purging approach may not fully solve endogeneity (demand may respond to past supply changes).
- The relatively high impact on GDP growth may reflect sample composition (several crisis and near-crisis cases) and possible spillovers of SME-targeted tightening to other sectors (households, large firms).

*Italic: Source—_wp1498 - 15.      For robustness, we also estimate an alternative specification that, instead of the*

### 28.      The impact of credit supply shocks faced by SMEs on macroeconomic variables may

### _wp1498 - 28.      The impact of credit supply shocks faced by SMEs on macroeconomic variables may

### Sample split and methodology
- The sample is split into two sub-samples according to the SMEs share in the non-financial business sector’s value added.
- The split threshold is the sample’s median: Median= 62 percent.
- Observations for Lithuania, the Netherlands, and Slovenia that crossed the 62 percent threshold over time were split between the two sub-samples.
- Breakdown of the two subsamples appears in Table 3A in the Annex.
- Table 3A composition:
  - High share: Cyprus, Greece, Italy, Spain, Portugal, Malta, Lithuania (2006q1–2012q4), the Netherlands (2006q1–2012q4), and Slovenia (2007q1–2012q4).
  - Low share: Austria, Germany, France, Hungary, Lithuania (2003q1–2005q4), the Netherlands (2003q1–2005q4), Poland, Slovenia (2003q1–2006q4), and the UK.
- VAR estimation and impulse response functions (IRFs) are used to evaluate the impact of adverse credit supply shocks applied to SMEs; IRFs are reported for High share and Low share sub-samples (Figures 4A and 5A in the Annex).

### Key empirical findings
- The prevalence of SMEs matters during periods of financial stress:
  - An adverse credit supply shock (standardized to be the same in both sub-samples) has a significant impact on GDP growth in both sub-samples.
  - The impact is much more prominent in the “High share” sub-sample (Figure 12).
- Quantified impacts (medium term, cumulative):
  - A credit supply shock (a “net tightening” by 18 percent of the banks) leads, on average, to a cumulative contraction of 2¾ percent in real output in the “High share” sub-sample.
  - The same shock leads to a cumulative contraction of 1¼ percent in real output in the “Low share” sub-sample.
- Suggested mechanism:
  - Stronger feedback effects from weak economic activity to banks’ balance sheets in High share countries may prolong tightening of lending standards and keep the interest rate spread higher for a longer period.
- Corroborating patterns:
  - The pace of economic recovery and credit growth during 2010–2012 is negatively correlated with the prevalence of SMEs across EU countries.
  - Countries with high share of SMEs tended to recover more slowly from the global financial crisis than countries with low share of SMEs.

### Robustness and caveats
- Results are statistically robust but should be treated with caution because:
  - The “High share” group consists of four countries that were at the epicenter of the crisis (Italy, Cyprus, Portugal, and Spain).
  - The impact of tightening lending conditions for SMEs on GDP growth may be overstated if it captures broader problems in domestic financial intermediation, tighter fiscal policies, and negative confidence effects.
  - Some high-prevalence SME countries are crisis or near-crisis cases where weak economic activity was driven by factors other than tight credit conditions.

### Policy implications and recommendations
- Maintain adequate access to finance for SMEs to support sustainable economic recovery, particularly in countries with high prevalence of SMEs.
- Differentiate policy support between:
  - High-productivity and solvent SMEs; and
  - SMEs that lack a sustainable business plan.
- Short- to medium-term measures that may have limited success:
  - Cheap long-term liquidity provision by central banks (examples in text: “Funding for Lending” by the Bank of England, and “Funding for Growth” by the National Bank of Hungary).
  - Expansion of state guarantee schemes, especially if conditional or linked to new lending to SMEs.
- Structural and financial intermediation measures needed:
  - Reduce fragmentation in credit markets and restore financial intermediation.
  - Strengthen banks’ balance sheets through a faster cleanup of bad assets and measures to ensure adequate capital and liquidity positions.
  - Increase availability of non-bank financing sources, including development of securitization markets for SMEs loans.

*Source: IMF staff estimations and analysis as presented in the supplied content.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp1498.pdf_