## wpiea2020084-print-pdf

## Source details

**Canonical URL:** [wpiea2020084-print-pdf](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020084-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020084-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020084-print-pdf.pdf.json)

---

### I. Introduction — context and research questions
- Public banks can:
  - Help resolve credit market inefficiencies (Eslava and Freixas(2016)).
  - Reduce the procyclicality of credit (Micco and Panizza(2006); Brei and Schclarek(2013); Cull and Martínez Pería(2013); Bertay, Demirgüç-Kunt and Huizinga(2015)).
  - Support projects with positive externalities (Behr, Foos and Norden(2017); Ogura(2018)).
  - Extend financial networks to a greater fraction of the population (Anson et al.(2013)).
- Potential downsides documented:
  - Sensitivity to political influences (Dinç (2005); Carvalho (2014)).
  - Inefficiency (Micco and Panizza(2006); Coleman and Feler(2015)).
  - Association with slower growth and financial development (La Porta, Lopez-de Silanes and Shleifer(2002)).
- Research focus:
  - Study how public banks in emerging markets (EMs) maintained lending during the Global Financial Crisis (GFC).
  - Three hypotheses examined:
    1. Greater lending due to sounder fundamentals.
    2. Greater lending due to funding advantages (deposits or state funds).
    3. [Hypothesis 3 text not included in excerpt.]

### Data and sample
- Quarterly bank-level dataset from central banks of 25 emerging market countries for 2006Q1-2010Q4.
- Initial sample: 1,062 firms (commercial banks, investment banks, savings banks, development banks, specialty banks, other financial entities).
- Final sample for analysis: 877 banks, of which 96 are public banks.
- Winsorization: all non-categorical variables winsorized at 5% in each quarter.
- Classification rules:
  - Foreign if branch of foreign-owned bank or majority-owned by foreign shareholders; otherwise domestic.
  - Domestic bank publicly owned if indicated in account filings or supplemented by Bankscope and online searches.
  - Commercial banks selected by “de facto” rule: exclude banks with average loans to assets ratio or deposits to non-equity liabilities ratio < 10% over sample period.
- Public bank asset share as of 2008Q4:
  - Range: 0% (Honduras, Jamaica, Nicaragua) to 73.56% (Belarus).
  - Average across all countries: 18.40%.
  - Regional averages: Latin American and Caribbean 11.61%; Eastern European and Central Asian 28.24%; East Asian 37.41%.

### Empirical methodology
- Crisis period indicator: 2008Q4-2009Q1.
- Baseline OLS specification: ∆log(loans)ijt with key regressors Publicijt, Publicijt × GFCt, Foreignijt, Foreignijt × GFCt.
- Controls Xij,t−1: share of total banking system assets, ratio of equity to assets, ratio of liquid assets, ratio of non-deposit liabilities to total liabilities; NPL ratio included in some specifications.
- Country-quarter fixed effects αjt to control for credit demand.
- Bank fundamentals lagged by one quarter; standard errors clustered by country-quarter.
- Additional methods:
  - Interactions of fundamentals with ownership and crisis indicators.
  - Semiparametric distribution regressions (Chernozhukov, Fernández-Val and Melly(2013)).
  - Difference-in-differences on NPLs using annual Bankscope data comparing 2008Q3 to mean 2012-2015.

### Findings — Hypothesis 1: Sounder fundamentals
- Descriptive: Before and during the crisis, public banks were significantly larger and had a greater asset share of liquid assets.
- Key regression results (Table 5 benchmarks):
  - Publicijt × GFCt positive and significant at 1% in country-quarter fixed effects regression: public banks increased lending by approximately 5% more than private banks during the crisis (conditional on country-quarter effects).
  - With controls, conditional effect of public ownership remains positive, significant, and of similar magnitude.
  - Including NPLs and interactions yields average public ownership effect during the crisis equal to 4.2% and significant at 5%.
- Interpretation: Differences in fundamentals do not fully explain greater lending by public banks during the GFC.

### Findings — Hypothesis 2: Funding advantages
- Deposits:
  - Public banks acquired about 2.4% more deposits than domestic private banks during the crisis (conditional on fundamentals and country-quarter effects).
  - Peak effect in 2008Q4: coefficient 4.9% and significant at 10%.
  - 2009Q1 coefficient 1.3% and insignificant.
  - Deposit pass-through: a 1% increase in deposits increases lending by about 0.25%.
  - Deposits growth does not explain increased lending: Publicijt × GFCt on loans growth remains significant when deposits growth and its interaction with the crisis are included.
- State funds:
  - Coefficient on Publicijt × GFCt is insignificant for state funds growth (limited observations where state funds > 0).
  - Quarterly differences in state funds: public ownership dummy positive and significant in non-crisis years, but interaction with crisis is insignificant.
- Asset composition:
  - Growth in share of liquid assets: negative and significant coefficient on Publicijt × GFCt indicates private banks increased liquid assets more than public banks during the crisis.
- Interpretation: Public banks attracted more deposits at the peak but not enough to explain lending differences; public banks did not obtain significantly more state funds during the crisis; public banks shifted asset composition away from liquid assets toward loans.

### Findings — Hypothesis 3: Stabilization motives
- Subsample regressions:
  - Public banks were on average less procyclical than private banks before the crisis.
  - Public banks were less procyclical during the crisis with notably larger magnitude compared to before the crisis.
  - After the crisis, the estimate is insignificant, suggesting public banks did not fully wind down lending after expanding during the crisis.
- Semiparametric distribution regressions (2007Q1-2007Q4; 2008Q4-2009Q1; 2010Q1-2010Q4):
  - During crisis, distribution of loans growth for public banks is uniformly greater across all quantiles compared to the counterfactual distribution (public banks’ characteristics mapped to private-bank behavior).
  - Before crisis and in 2010, distributions are similar except among fastest-growing banks (above the 80th percentile).
  - Omitting top 20% of observations dramatically diminishes negative effect of public ownership before the GFC, indicating mean estimates may be driven by outliers.
- Interpretation: Countercyclical behavior of public banks was unique to the crisis and consistent with an objective to stabilize credit supply rather than a persistent business model difference across normal times.

### Long-term implications and performance
- Difference-in-differences on NPLs comparing 2008Q3 to mean 2012-2015:
  - NPLs of public banks did not significantly change relative to private banks.
- Caveats:
  - NPLs are reported measures; analysis assumes no ownership-correlated misreporting.
  - Non-significant changes in NPLs do not rule out other inefficiencies, political targeting, or low productivity in public banks’ lending (e.g., Brazil case studies).
- Interpretation: Public banks’ countercyclical lending during the GFC did not appear to compromise stability as measured by reported NPL ratios.

### Key regression estimates and sample statistics (selected, exact values preserved)
- Main loans growth regression (Table 5, Baseline column (2)):
  - Public: -0.607* (t-statistic: -1.91)
  - Public × GFC: 5.605*** (4.71)
  - Foreign: -1.066*** (-3.66)
  - Foreign × GFC: 3.012*** (2.75)
  - Lag asset share: -0.127*** (-6.29)
  - Lag capital ratio: 0.070*** (5.07)
  - Lag liquidity ratio: 0.064*** (6.28)
  - R2 = 0.259; Observations = 14,461
- Deposits and loans regressions (Table 7, selected):
  - Deposits growth: Public × GFC (2008Q4) = 3.272* (1.86)
  - Loans growth: Deposits growth coefficient = 0.250*** (22.85)
  - Loans growth: Public × GFC = 4.716*** (4.26)
  - Observations for loans regression = 14,414; R2 = 0.334
- Subsample loans growth (Table 9):
  - 2006Q1-2008Q3: Public = -1.236*** (-2.81); Observations = 7,611
  - 2008Q4-2009Q1: Public = 4.435*** (3.99); Observations = 1,545
  - 2009Q2-2010Q4: Public = 0.354 (0.83); Observations = 5,308
- Sample summary statistics (Table 2, exact values):
  - Loans growth: N = 14,504; Mean = 6.142; SDSD = 4.370; 12.037
  - Deposits growth: N = 14,471; Mean = 6.091; SDSD = 4.130; 14.832
  - State funds growth: N = 9,277; Mean = 0.054; SDSD = 0.000; 0.414
  - Public: N = 15,479; Mean = 0.111; SDSD = 0.000; 0.314
  - Foreign: N = 15,479; Mean = 0.369; SDSD = 0.000; 0.483
  - Asset share: N = 15,479; Mean = 2.843; SDSD = 0.766; 4.270
  - Capital ratio: N = 15,477; Mean = 15.416; SDSD = 10.963; 11.956
  - Liquidity ratio: N = 15,439; Mean = 28.177; SDSD = 25.358; 18.484
  - Wholesale ratio: N = 15,472; Mean = 32.990; SDSD = 25.889; 25.219
  - NPL ratio: N = 12,720; Mean = 3.016; SDSD = 1.720; 3.471

### Conclusions and policy implications
- Core conclusions:
  - Public bank lending during the GFC cannot be explained solely by sounder fundamentals or superior access to deposits or state funds.
  - Evidence is consistent with public banks lending more because they pursued an objective of helping to stabilize the economy during the crisis.
  - Public banks achieved stabilization partly by shifting asset composition from liquid instruments to loans.
  - The countercyclical lending behavior was specific to the crisis period rather than a regular pre- or post-crisis characteristic.
  - No significant deterioration in reported NPLs for public banks in the post-crisis period was found.
- Policy relevance:
  - Public banks can play a stabilizing role during crises, helping to avoid a credit crunch and potentially mitigating persistent GDP losses from hysteresis.
  - Risks from inefficiencies and political influences in public banks during normal times remain; these challenges could outweigh crisis-time benefits if not continuously addressed.

*Source: wpiea2020084-print-pdf*

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

### I. INTRODUCTION

### Context and debates on public banks
- Public banks can help resolve credit market inefficiencies (Eslava and Freixas(2016)).
- Public banks can reduce the procyclicality of credit (Micco and Panizza(2006), Brei and Schclarek(2013), Cull and Martínez Pería(2013), and Bertay, Demirgüç-Kunt and Huizinga(2015)).
- Public banks can support projects with positive externalities (e.g., lending to small and medium-sized enterprises as highlighted by Behr, Foos and Norden(2017) and Ogura(2018)).
- Public banks can extend financial networks to a greater fraction of the population (Anson et al.(2013)).
- Potential downsides noted:
  - Sensitivity to political influences (Dinç (2005), Carvalho (2014)).
  - Inefficiency (Micco and Panizza(2006), Coleman and Feler(2015)).
  - Association with slower growth and financial development (La Porta, Lopez-de Silanes and Shleifer(2002)).

### Observations from the Global Financial Crisis (GFC)
- During the Global Financial Crisis (GFC), public banks experienced little contraction in lending compared to domestic private banks in some countries (see Figure1 as well as Brei and Schclarek (2013), Cull and Martínez Pería(2013), and Bertay, Demirgüç-Kunt and Huizinga(2015)), which suggests that they can also help stabilize the economy by avoiding a credit crunch.
- Note on data limitations: cross-country studies typically use annual data, which is unsuitable for detecting short-lived depositor behavior as well as the fast-changing features of the GFC. Previously used datasets also lack sufficiently detailed balance-sheet breakdowns, such as measures of the funding from the public sector.

### Research focus and hypotheses of this paper
- This paper studies how public banks in emerging markets (EMs) maintained lending during the GFC.
- Three hypotheses considered:
  1. Public banks lent more because they exhibited sounder fundamentals that enabled them to take greater risks.
  2. Public banks lent more because they benefitted from funding advantages, such as safe haven perceptions of depositors or special access to government funding.
  3. [Text for hypothesis 3 not included in the supplied excerpt.]

### Supporting country-specific evidence (footnote)
- Some country specific studies have also highlighted similar results. For example, Coleman and Feler(2015) shows that localities in Brazil with a high share of public banks received more loans and experienced better employment outcomes relative to localities with a low share of government banks.
- In the case of Turkey, Önder and Özyıldırım(2013) show that credit provided by public banks during the crisis has a significant and positive effect on local growth.

*Source: wpiea2020084-print-pdf - References .............................................................................................................*

### 3. Public banks lent more because they pursued an objective of helping stabilize the economy

### 3. Public banks lent more because they pursued an objective of helping stabilize the economy

### Data and sample
- Quarterly bank-level dataset from central banks of 25 emerging market (EM) countries for the period 2006Q1-2010Q4.
- Initial dataset: 1,062 firms (commercial banks, investment banks, savings banks, development banks, specialty banks, and other financial entities).
- Final sample for analysis: 877 banks, of which 96 are public banks.
- All non-categorical variables are winsorized at 5% in each quarter.
- Classification rules:
  - A bank is foreign if it is a branch office of a foreign-owned bank or majority-owned by foreign shareholders; otherwise it is domestic.
  - A domestic bank is publicly owned if indicated in account filings or supplemented by Bankscope and online searches.
  - Commercial banks selected by a “de facto” rule: exclude banks whose average loans to assets ratio or deposits to non-equity liabilities ratio is less than 10% over the sample period.
- Public bank asset share as of 2008Q4:
  - Range: 0% (Honduras, Jamaica, Nicaragua) to 73.56% (Belarus).
  - Average across all countries: 18.40%.
  - Regional averages: Latin American and Caribbean 11.61%; Eastern European and Central Asian 28.24%; East Asian 37.41%.

### Empirical methodology
- Crisis period indicator: 2008Q4-2009Q1.
- Baseline OLS specification (loans growth ∆log(loans)ijt):
  - Key regressors: Publicijt, Publicijt × GFCt, Foreignijt, Foreignijt × GFCt.
  - Controls Xij,t−1: share of total banking system assets, ratio of equity to assets, ratio of liquid assets, ratio of non-deposit liabilities to total liabilities; NPL ratio included in some specifications.
  - Country-quarter fixed effects αjt to control for credit demand.
  - Bank fundamentals lagged by one quarter; standard errors clustered by country-quarter.
- Additional methods:
  - Interactions of fundamentals with public and crisis indicators to assess heterogeneous sensitivities.
  - Semiparametric distribution regressions (Chernozhukov, Fernández-Val and Melly(2013)) to compare distributions of loan growth for public and private banks and counterfactuals.
  - Difference-in-differences on NPLs using annual Bankscope data comparing 2008Q3 to mean 2012-2015.

### Findings — Hypothesis 1: Sounder fundamentals
- Before and during the crisis, public banks were significantly larger and had a greater asset share of liquid assets.
- Key regression results (Table 5 benchmark):
  - Publicijt × GFCt positive and significant at 1% in a country-quarter fixed effects regression: public banks increased lending by approximately 5% more than private banks during the crisis (conditional on country-quarter effects).
  - With control variables included, the conditional effect of public ownership remains positive, significant, and of similar magnitude.
  - Including NPLs and interacting fundamentals with ownership and crisis indicators yields an average public ownership effect during the crisis equal to 4.2% and significant at 5%.
- Interpretation: Differences in fundamentals do not fully explain the greater lending by public banks during the GFC.

### Findings — Hypothesis 2: Funding advantages
- Deposits:
  - Regression of deposits growth shows public banks acquired about 2.4% more deposits than domestic private banks during the crisis (conditional on fundamentals and country-quarter effects).
  - Peak effect in 2008Q4: coefficient 4.9% and significant at 10%.
  - 2009Q1 coefficient 1.3% and insignificant.
  - Deposit pass-through: a 1% increase in deposits increases lending by about 0.25%.
  - Deposits growth does not explain the increased lending of public banks during the crisis: Publicijt × GFCt on loans growth remains significant when deposits growth and its interaction with the crisis are included.
- State funds:
  - Regressions with state funds growth as dependent variable show the coefficient on Publicijt × GFCt is insignificant (limited observations where state funds > 0).
  - Using quarterly differences in state funds: public ownership dummy positive and significant in non-crisis years, but the interaction with the crisis is insignificant.
- Asset composition:
  - Regression on growth in the share of liquid assets: negative and significant coefficient on Publicijt × GFCt indicates private banks increased liquid assets more than public banks during the crisis.
- Interpretation: Public banks attracted more deposits at the peak but not enough to explain lending differences; public banks did not obtain significantly more state funds during the crisis; public banks shifted asset composition away from liquid assets toward loans.

### Findings — Hypothesis 3: Stabilization motives
- Subsample regressions before, during, and after the crisis:
  - Public banks were on average less procyclical than private banks before the crisis.
  - Public banks were less procyclical during the crisis with a notably larger magnitude compared to before the crisis.
  - After the crisis, the estimate is insignificant, suggesting public banks did not fully wind down lending after expanding during the crisis.
- Semiparametric distribution regressions (2007Q1-2007Q4, 2008Q4-2009Q1, 2010Q1-2010Q4):
  - During the crisis, the distribution of loans growth for public banks is uniformly greater across all quantiles compared to the counterfactual distribution (public banks’ characteristics mapped to private-bank behavior).
  - Before the crisis and in 2010, distributions are similar except among fastest-growing banks (above the 80th percentile).
  - Omitting the top 20% of observations dramatically diminishes the negative effect of public ownership before the GFC, indicating mean estimates may be driven by outliers.
- Interpretation: The countercyclical behavior of public banks was unique to the crisis and consistent with an objective to stabilize credit supply rather than a persistent business model difference across normal times.

### Long-term implications and performance
- Difference-in-differences on NPLs comparing pre-crisis (2008Q3) to mean 2012-2015:
  - NPLs of public banks did not significantly change relative to private banks.
- Caveats:
  - NPLs are reported measures; analysis assumes no ownership-correlated misreporting.
  - Non-significant changes in NPLs do not rule out other inefficiencies, political targeting, or low productivity in public banks’ lending (e.g., Brazil case studies).
- Interpretation: Public banks’ countercyclical lending during the GFC did not appear to compromise stability as measured by reported NPL ratios.

### Conclusions and policy implications
- Summary of main conclusions:
  - Public bank lending during the GFC cannot be explained solely by sounder fundamentals or superior access to deposits or state funds.
  - Evidence is consistent with public banks lending more because they pursued an objective of helping to stabilize the economy during the crisis.
  - Public banks achieved stabilization partly by shifting asset composition from liquid instruments to loans.
  - The countercyclical lending behavior was specific to the crisis period rather than a regular pre- or post-crisis characteristic.
  - No significant deterioration in reported NPLs for public banks in the post-crisis period was found.
- Policy relevance:
  - Public banks can play a stabilizing role during crises, helping to avoid a credit crunch and potentially mitigating persistent GDP losses from hysteresis.
  - Risks remain from documented inefficiencies and political influences in public banks during normal times; these challenges could outweigh crisis-time benefits if not continuously addressed.

*Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020084-print-pdf.pdf*

### APPENDIX I.FIGURES

### APPENDIX I.FIGURES

### Figures: loans growth patterns and distributions
- Figure 1: Mean quarterly loans growth of domestic private and domestic public banks within a window of the GFC for sample banks from 25 emerging market countries (date labels shown: 2006q2, 2007q3, 2008q4, 2010q1, 2011q2). Vertical axis: Quarterly loans growth (%).
- Figure 2: Empirical distributions of loans growth:
  - Panel (a) Before GFC (2007Q1-2007Q4): observed private quantiles, observed public quantiles, counterfactual quantiles.
  - Panel (b) During GFC (2008Q4-2009Q1): observed private quantiles, observed public quantiles, counterfactual quantiles.
  - Panel (c) After GFC (2010Q1-2010Q4): observed private quantiles, observed public quantiles, counterfactual quantiles.

### Variable definitions (Table 1)
- Public: Indicator of public bank (Bankscope, online sources).
- Foreign: Indicator of foreign bank (Claessens and van Horen (2015), IMF Bank Contagion Module, Bankscope, online sources).
- Loans growth: Quarterly growth rate of loans (= net loans when available) (%), source: Central banks.
- Deposits growth: Quarterly growth rate of customer deposits (= demand + time + savings deposits, when available) (%), source: Central banks.
- State funds growth: Quarterly growth rate of state funds (= central bank deposits and other public sector liabilities) (%), source: Central banks.
- Asset share: Share of assets in the banking system (%), source: Central banks.
- Capital ratio: Equity divided by assets (%), source: Central banks.
- Liquidity ratio: Liquid assets (= cash + securities) divided by total assets (%), source: Central banks.
- Wholesale ratio: (Non-equity liabilities - customer deposits)/non-equity liabilities (%), source: Central banks.
- Non-performing loans (NPL): Non-performing loans divided by gross loans (%), source: Central banks and Bankscope.

### Sample summary statistics (Table 2)
- Loans growth: N = 14,504; Mean = 6.142; SDSD = 4.370; 12.037
- Deposits growth: N = 14,471; Mean = 6.091; SDSD = 4.130; 14.832
- State funds growth: N = 9,277; Mean = 0.054; SDSD = 0.000; 0.414
- Public: N = 15,479; Mean = 0.111; SDSD = 0.000; 0.314
- Foreign: N = 15,479; Mean = 0.369; SDSD = 0.000; 0.483
- Asset share: N = 15,479; Mean = 2.843; SDSD = 0.766; 4.270
- Capital ratio: N = 15,477; Mean = 15.416; SDSD = 10.963; 11.956
- Liquidity ratio: N = 15,439; Mean = 28.177; SDSD = 25.358; 18.484
- Wholesale ratio: N = 15,472; Mean = 32.990; SDSD = 25.889; 25.219
- Non-performing loans ratio (NPL): N = 12,720; Mean = 3.016; SDSD = 1.720; 3.471

### Public bank presence by country (Table 3, values as of 2008Q3)
- Selected entries (Country / Total banks / Public banks / Public share):
  - Argentina: 62 / 11 / 33.891
  - Belarus: 29 / 4 / 73.56
  - Bolivia: 11 / 6 / 856
  - Brazil: 97 / 12 / 34.139
  - Chile: 21 / 11 / 3.375
  - Indonesia: 117 / 53 / 6.644
  - Taiwan Province of China: 379 / 59 / 9.419
  - Turkey: 263 / 25 / 25.304
  - Venezuela: 49 / 7 / 12.091
- (Full table in source lists all sample countries with total banks, public banks, and public share.)

### Fundamentals comparison (Table 4)
- Means of bank-level averages before (2006Q1-2008Q3), during (2008Q4-2009Q1), and after (2009Q2-2010Q4) the crisis for domestic private and public banks; p-values from difference of means tests reported.
- Before GFC (Private / Public / P-Value):
  - Asset share: 2.214 / 4.616 / 0
  - Capital ratio: 15.17 / 13.89 / 0.313
  - Liquidity ratio: 25.86 / 35.60 / 0
  - Wholesale ratio: 28.95 / 32.90 / 0.131
  - NPL ratio: 3.21 / 4.25 / 0.016
- During GFC:
  - Asset share: 2.232 / 4.90 / 0
  - Capital ratio: 16.03 / 14.30 / 0.255
  - Liquidity ratio: 26.17 / 33.62 / 0
  - Wholesale ratio: 31.08 / 35.58 / 0.129
  - NPL ratio: 3.42 / 3.53 / 0.805
- After GFC:
  - Asset share: 2.276 / 5.29 / 0
  - Capital ratio: 15.00 / 13.04 / 0.132
  - Liquidity ratio: 28.15 / 33.88 / 0.004
  - Wholesale ratio: 29.28 / 36.20 / 0.010
  - NPL ratio: 3.26 / 3.80 / 0.187

### Main loans growth regressions (Table 5)
- Estimating equation (1); country-quarter-clustered standard errors; significance levels indicated.
- Key coefficient estimates (Baseline column (2)):
  - Public: -0.607* (t-statistic in parentheses: -1.91)
  - Public × GFC: 5.605*** (4.71)
  - Foreign: -1.066*** (-3.66)
  - Foreign × GFC: 3.012*** (2.75)
  - Lag asset share: -0.127*** (-6.29)
  - Lag capital ratio: 0.070*** (5.07)
  - Lag liquidity ratio: 0.064*** (6.28)
  - R2 = 0.259; Observations = 14,461; Country-quarter FE: Yes

- No controls (column (1)):
  - Public: -0.640** (-2.01)
  - Public × GFC: 5.639*** (4.59)
  - Foreign: -0.937*** (-3.14)
  - Foreign × GFC: 3.056*** (2.80)
  - Observations = 14,504; R2 = 0.248

- Additional controls (column (3)):
  - Public: -0.347 (-1.02)
  - Public × GFC: 5.788*** (5.50)
  - Foreign: -1.226*** (-3.75)
  - Foreign × GFC: 1.981* (1.85)
  - Lag NPL ratio: -0.432*** (-8.67)
  - Observations = 11,890; R2 = 0.267

### Regressions with rich interactions (Table 6)
- Loans growth regression with interactions between fundamentals, public indicator, GFC indicator, and triple interactions.
- Selected coefficients:
  - Public: -0.722 (-0.79)
  - Public × GFC: 6.275** (2.35)
  - Foreign: -1.069*** (-3.67)
  - Foreign × GFC: 2.758*** (2.64)
  - Lag asset share: -0.124*** (-5.43)
  - Lag capital ratio: 0.074*** (4.76)
  - Lag liquidity ratio: 0.059*** (5.09)
  - Observations = 14,464; R2 = 0.260; Country-quarter FE: Yes

### Deposits and loans regressions (Table 7)
- Dependent variable: Deposits growth in columns (1)-(3); Loans growth in column (4).
- Selected estimates (column (1), Y = Deposits growth):
  - Public: -0.844** (-2.07)
  - Public × GFC (GFC=2008Q4): 3.272* (1.86)
  - Foreign: -1.355*** (-4.22)
  - Foreign × GFC: 3.772** (2.37)
  - Lag capital ratio: 0.133*** (7.57)
  - Lag liquidity ratio: -0.042*** (-3.38)
  - Lag wholesale ratio: 0.090*** (10.46)
  - Observations = 14,430; R2 = 0.220; Country-quarter FE: Yes

- Column (4), Y = Loans growth:
  - Public: -0.408 (-1.28)
  - Public × GFC: 4.716*** (4.26)
  - Foreign: -0.718*** (-2.69)
  - Foreign × GFC: 1.975** (2.04)
  - Deposits growth coefficient: 0.250*** (22.85)
  - Observations = 14,414; R2 = 0.334

### Additional explanations: state funds, liquidity (Table 8)
- Dependent variables: state funds growth (col 1), quarterly difference in state funds (col 2), liquidity ratio growth (col 3).
- Selected estimates:
  - Public (col 2): 0.091*** (4.03)
  - Public × GFC (col 3): -4.719* (-1.89)
  - Foreign × GFC (col 1): 58.709 (0.70)
  - Lag liquidity ratio (col 3): -0.385*** (-11.87)
  - Observations: col1 = 239; col2 = 19,271; col3 = 14,290
  - R2: col1 = 0.367; col2 = 0.321; col3 = 0.176

### Subsample and robustness results (Tables 9 and 10)
- Loans growth subsamples (Table 9):
  - 2006Q1-2008Q3:
    - Public: -1.236*** (-2.81)
    - Foreign: -0.786** (-2.03)
    - Observations = 7,611; R2 = 0.227
  - 2008Q4-2009Q1:
    - Public: 4.435*** (3.99)
    - Foreign: 1.757* (1.71)
    - Observations = 1,545; R2 = 0.319
  - 2009Q2-2010Q4:
    - Public: 0.354 (0.83)
    - Foreign: -1.295*** (-3.04)
    - Observations = 5,308; R2 = 0.193

- Omitting fast-growing banks (Table 10) — omitting banks with loans growth above the 80th percentile by country, quarter, and ownership type:
  - 2006Q1-2008Q3:
    - Public: -0.447 (-1.25)
    - Observations = 5,876; R2 = 0.349
  - 2008Q4-2009Q1:
    - Public: 4.786*** (3.45)
    - Observations = 1,193; R2 = 0.514
  - 2009Q2-2010Q4:
    - Public: 0.606 (1.38)
    - Observations = 4,100; R2 = 0.348

### Long-term effects (Table 11)
- Estimating equation (2); dependent variable: NPL ratio; heteroscedasticity-robust standard errors.
- Coefficients:
  - Post-GFC: 0.119 (0.22)
  - Public: 0.391 (0.37)
  - Post-GFC × Public: 0.345 (0.23)
  - Observations = 588; R2 = 0.264; Country-quarter FE: Yes

*Source: wpiea2020084-print-pdf - APPENDIX I.FIGURES*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020084-print-pdf.pdf_