## wpiea2021007-print-pdf — Section 1–4 summary

## Source details

**Canonical URL:** [wpiea2021007-print-pdf — Section 1–4 summary](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021007-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021007-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021007-print-pdf.pdf.json)

---

### Abstract and central finding
- Scope: cross-country bank-level data for 125 advanced and developing economies for 1999–2018.
- Main findings:
  - Public bank lending is less procyclical than private bank lending on average, particularly during busts.
  - In developing economies with high public debt, public bank lending has been more procyclical, particularly outside of the global financial crisis (GFC) period.
  - High public debt likely raises financing costs for public banks and reduces subsidies they receive in economies with tighter budget constraints.
- Policy implication (summary): Public banks can contribute to stabilization if financially sound and well resourced, especially in economies with low public debt, but potential benefits must be weighed against economic and fiscal costs including misallocation of credit and contingent liabilities.

### Data, definitions, and sample characteristics
- Primary data sources:
  - Fitch Connect bank-level balance sheets and income statements.
  - October 2019 vintage of the IMF’s World Economic Outlook database.
  - IMF’s Financial Development database.
- Sample after cleaning:
  - More than 4,000 (3,000) banks from 45 advanced and 80 (75) developing economies for 1999–2018 (2010–18).
  - Around 7 (8.5) percent of banks in the sample are public.
- Public bank identification rules:
  - Reported as “government sponsored enterprises” or “public entities”, or state ownership share over 25 percent.
  - National development banks treated as public banks in baseline (relaxing assumption does not change main findings qualitatively).
  - Alternative definition in robustness: over 50 percent government ownership (majority).
- Exclusions and cleaning:
  - Banks with total assets under US$5 billion excluded.
  - Eliminated banks with total equity or total employees less than zero; ratios capped (for example, deposits over total liabilities between 0% and 100%); returns between -100% and 100%; remaining indicators restricted to 1st–99th percentiles.
- Public debt high/low definition:
  - High public debt = general government debt above the 75th percentile of the sample distribution.
  - Roughly 100 percent of GDP for advanced economies and roughly 60 percent of GDP for developing economies.

### Empirical methodology
- Estimation: panel fixed effects regressions (bank and year fixed effects).
- Dependent variable: growth rate of net loans in current US dollars for bank j in country i, year t.
- Baseline cyclical indicator: growth rate of GDP per capita relative to its average growth rate in the past 20 years (and alternative: GDP per capita growth rate).
- Key interactions:
  - Public bank dummy * cycle.
  - Public bank * cycle * high public debt dummy.
- Controls:
  - Lagged dependent variable and lagged growth of net loans.
  - Lagged bank-level characteristics: log total assets, equity/total assets, liquid assets/total assets, return on average assets, deposits/total liabilities, net loans/total assets, nonperforming/gross loans.
  - Level of financial development and year fixed effects.
- Robustness: results robust to defining public banks as majority government owned, dropping banks without eight years of consecutive data between 2005 and 2012, treating development banks differently, and using local currency loan growth.

### Core empirical findings (aggregate and interpreted magnitudes)
- Private banks:
  - Private bank net lending growth increases by 2 (-0.4) percentage points in developing (advanced) economies in response to a 1 percentage point increase in the growth rate of GDP per capita relative to its trend (decade after the GFC, columns (11) and (12)).
- Public banks relative to private banks:
  - Public banks’ net lending grows 1.5 (3.8) percentage points less in developing (advanced) economies with low public debt levels in the decade after the GFC (columns (11) and (12)), yielding less procyclical (or countercyclical) lending by public banks in developing (advanced) economies with low public debt levels.
  - When public debt is high, public bank net lending grows 2.5 (2.7) percentage points more in developing (advanced) economies compared to when public debt is low; the difference is statistically significant only for developing economies.
- Heterogeneity by public debt and crisis period:
  - Public banks are less procyclical in economies with low public debt (negative coefficients for “public bank * cycle” in columns (5) to (12)).
  - In economies with high public debt, particularly in developing economies and outside the GFC period, public banks are not less procyclical; coefficients for “public bank * cycle * high public debt” are positive in columns (7) to (12).
  - Distinguishing booms and busts:
    - Private bank lending is procyclical during both booms and busts in developing economies.
    - In advanced economies, private bank lending is procyclical during booms and countercyclical during busts.
    - Public bank lending is less procyclical on average, particularly during busts, and in economies with low public debt.
    - In economies with high public debt, public bank lending is procyclical during booms, and during busts in developing economies particularly after the GFC.

### Mechanisms and illustrative cases
- Mechanism: High public debt raises external borrowing costs and tightens budget constraints, increasing financing costs for public banks, reducing government subsidies, and in some cases prompting governments to draw funds from public banks (for example, through increased dividends).
- Exception: Global crises like the GFC can be “exceptional” — sovereign and bank borrowing costs in developing economies may remain favorable amid low global interest rates and loose monetary policy in advanced economies, enabling public banks to expand lending during global crises.
- Case examples reported:
  - Brazil (BNDES): lending surged during the GFC and recovery following a large capital injection, but declined sharply and procyclically during the domestic recession of 2014–16 once soaring public deficits and debt constrained policy lending.
  - India: government guarantees allowed public banks to expand credit during the GFC but led to deterioration in loan quality of vulnerable public banks and increased overall financial sector fragility.

### Selected reported coefficient patterns (exact values preserved as in source tables)
- Representative cyclical coefficients and interactions (examples reported in tables):
  - GDP pc Growth Gap (Cycle) examples: 0.226; -0.181; 0.537***; 0.595**; 1.611***.
  - Public Bank * Cycle examples: -0.842; -1.398***; -1.204**; -2.816***; -3.345***; -4.253**.
  - Public Bank * Cycle * HD examples: 1.319; 3.185; 1.629*; 8.411; 4.765***.
  - Log of Total Assets = L examples: -0.172***; -0.200***; -0.195***; -0.324***; -0.359***.
  - Net Loans / Total Assets = L examples: -0.454***; -0.428***; -0.423***; -0.979***; -0.961***.
  - Nonperforming / Gross Loans = L examples: -0.517***; -0.690***; -0.660***; -0.744***; -1.076***.
  - High Public Debt (HD) examples: 0.0192**; 0.0751***; -0.0531*; -0.0413.
- Table 9 (differentiating public development and commercial banks) selected interactions:
  - Public Development Bank * Cycle: -0.875; -1.987***; -1.866**; -2.375***; -4.279**.
  - Public Commercial Bank * Cycle: -0.0631; -0.729**; -1.660**; -4.140***.
  - Public Development Bank * Cycle * HD: 4.300***; 2.513***; 15.55***.
  - Public Commercial Bank * Cycle * HD: -2.335; 1.752*; 6.039***.

### Control variables and robustness
- Control variable patterns:
  - Net lending grows faster in developing economies with higher financial development (Financial Development = L sometimes positive and significant).
  - Faster lending associated with lower lagged loan size, lower bank size (total assets), higher capitalization rate (particularly in advanced economies), lower nonperforming loan ratios, higher profitability (particularly in developing economies), higher deposit funding ratios in developing economies, and lower liquidity ratios in developing economies.
- Robustness checks:
  - Using growth in constant local currency leaves main results qualitatively unchanged.
  - Dropping banks without eight years of consecutive data between 2005 and 2012, relaxing national development bank assumption, and defining public banks as majority government owned do not change main findings materially.
  - Alternative cyclical indicator (GDP per capita growth instead of gap) yields similar patterns.

### Interpretation, risks, and policy trade-offs
- Conditional stabilization potential:
  - Public banks can be part of a macroeconomic stabilization toolkit when public finances are strong and public banks are financially sound and well resourced.
- Constraints and risks:
  - High public debt constrains the countercyclical ability of public banks, especially during domestic busts in developing economies.
  - Potential adverse outcomes: misallocation of credit, crowding out of private credit, deterioration of loan quality in vulnerable banks, increased financial sector fragility, and realization of contingent fiscal liabilities (for example from extensive government guarantees).
  - Public banks can contribute to a sovereign-bank nexus where funding government needs and receiving government support create a sovereign-bank doom loop.
- Policy considerations and recommendations:
  - Weigh stabilization benefits of public banks against fiscal and economic costs.
  - Maintain public banks’ financial soundness and ensure adequate fiscal space before deploying countercyclical public bank interventions.
  - Strengthen governance and oversight: ensure transparency, clarify rationale for state intervention, assess and disclose risks and costs, implement risk mitigation measures, and make budgetary provisions for materialized risks.
  - Suggested further research: analyze role of government guarantees in boosting banks’ ability to lend during crises and whether such guarantees are disproportionately utilized by public banks.

*Source: WP/21/7 — Revisiting the Stabilization Role of Public Banks: Public Debt Matters, H. Elif Ture, January 2021 (Sections 1–4 of wpiea2021007-print-pdf).*

### Section 1

### wpiea2021007-print-pdf - Section 1

### Abstract and central finding
- Paper examines whether weak public finances hinder the stabilization role of public banks.
- Sample: cross-country bank-level data for 125 advanced and developing economies for 1999–2018.
- Main findings:
  - Public bank lending is less procyclical than private bank lending on average, particularly during busts.
  - In developing economies with high public debt levels, public bank lending has been more procyclical, particularly outside of the global financial crisis (GFC) period.
  - High public debt likely raises financing costs for public banks and reduces subsidies they receive in economies with tighter budget constraints.
- Policy implication (summary): Public banks can contribute to stabilization if financially sound and well resourced, especially in economies with low public debt, but potential benefits must be weighed against economic and fiscal costs including misallocation of credit and contingent liabilities.

### Background and motivation
- Public banks were widely used to counter the private credit crunch during the global financial crisis of 2008–09 and were active during the COVID-19 crisis.
- Literature context:
  - Previous studies show public bank lending less procyclical in developing economies and even countercyclical in developed economies.
  - This paper revisits those findings and incorporates the role of public debt.
- Note on scope: The analysis focuses on 1999–2018 when data coverage from Fitch Connect is better.

### Data sources, definitions, and sample characteristics
- Primary data source: Fitch Connect bank-level balance sheets and income statements.
- Country-level macro-fiscal data: October 2019 vintage of the IMF’s World Economic Outlook database.
- Financial development index: IMF’s Financial Development database.
- Identification rules:
  - Public banks: reported as “government sponsored enterprises” or “public entities” in the dataset, or those with state ownership share of over 25 percent.
  - National development banks (many lacking ownership data) are identified as public banks because governments often substantially influence their operations and funding.
  - Banks with total assets under US$5 billion are excluded.
  - Only latest annual 12-month financial statements used; nonconsolidated data used when available; banks without at least two years of consecutive data dropped.
- Sample after cleaning:
  - More than 4,000 (3,000) banks from 45 advanced and 80 (75) developing economies for 1999–2018 (2010–18).
  - Around 7 (8.5) percent of banks in the sample are public.
- Data cleaning rules included eliminating banks with total equity or total employees less than zero; limiting ratios (for example, deposits over total liabilities) to range between 0% and 100%; financial returns to range between -100% and 100%; and remaining indicators to lie within the 1st and 99th percentiles.
- Typical bank-level characteristics (median bank averages over time reported in Table 1):
  - Public banks tend to have more assets than private banks, especially in advanced economies.
  - Public banks have more net loans as a share of assets in developing economies, higher nonperforming loan ratios (particularly in developing economies), lower capitalization rates, lower deposit funding ratios in advanced economies, and lower profitability measured by return on average assets.

### Empirical methodology
- Estimation approach: panel fixed effects regressions (bank and year fixed effects).
- Dependent variable: growth rate of net loans in current US dollars for bank j in country i, year t.
- Baseline cyclical indicator: growth rate of GDP per capita relative to its average growth rate in the past 20 years.
- Key interactions:
  - Public bank dummy interacted with the cyclical indicator.
  - High public debt dummy interacted (economies with general government debt above the 75th percentile of the sample distribution).
- Definition of high public debt (75th percentile thresholds reported in the paper):
  - Roughly 100 percent of GDP for advanced economies.
  - Roughly 60 percent of GDP for developing economies.
- Controls include:
  - Lagged dependent variable (to capture persistence), lagged growth of net loans, lagged bank-level characteristics (bank size = log total assets, capitalization = equity over assets, liquidity = liquid over total assets, profitability = return on average assets, funding = deposits over liabilities, loan size = net loans over assets, loan quality = nonperforming loans over gross loans).
  - Level of financial development and year fixed effects.
- Standard errors robustified for autocorrelation and heteroskedasticity.
- Robustness checks mentioned:
  - Using majority government ownership to define public banks does not change main findings qualitatively.
  - Dropping banks without at least eight years of consecutive data between 2005 and 2012 (to address survival bias) does not change main findings qualitatively.
  - Treating national development banks as public banks is a modeling assumption; relaxing it does not change findings qualitatively.

### Key empirical results and interpretation
- Aggregate results:
  - Private bank lending is procyclical in developing economies.
  - In advanced economies, private bank lending is procyclical during booms but countercyclical during busts.
  - Public bank lending is less procyclical than private bank lending on average, with the difference most pronounced during busts and in economies with low public debt.
- High public debt interaction:
  - In developing economies with high public debt (above the 75th percentile, roughly 60 percent of GDP), public bank lending is more procyclical, particularly outside of the GFC period.
  - Mechanism: High public debt raises external borrowing costs and tightens budget constraints, leading to higher financing costs for public banks and lower government subsidies; governments may even draw funds from public banks (for example, through increased dividends).
- Exception for global crises:
  - Global busts like the GFC are "exceptional" because sovereign and bank borrowing costs in developing economies may remain favorable given low global interest rates and loose monetary policy in advanced economies; this can enable public banks to expand lending during global crises.
- Case studies / illustrative examples:
  - Brazil (BNDES):
    - Brazil is identified as a developing economy with high public debt in the sample (public debt above 75th percentile for developing economies ≈ 60 percent of GDP).
    - BNDES lending surged during the GFC and for a few years in the post-crisis recovery following a large capital injection by the state, but declined sharply and procyclically during the domestic recession of 2014–16 once soaring public deficits and debt closed the door on policy lending to public banks.
    - Studies indicate subsidized credit during the GFC largely benefited large firms without credit constraints and may have crowded out private credit, with subsequent instances of politically motivated lending during recovery.
  - India:
    - Government guarantees allowed public banks to expand credit during the GFC, but this led to deterioration in loan quality of vulnerable public banks and increased overall financial sector fragility.
- Overall interpretation:
  - Public banks can stabilize credit cycles when public finances are strong and public banks are financially sound and well resourced.
  - High public debt limits the stabilization role of public banks during domestic busts in developing economies because of higher borrowing costs, reduced subsidies, and potential fiscal strain that can reverse prior policy lending.

### Conclusions and policy implications (from Section I and concluding remarks)
- Public banks have conditional stabilization potential:
  - They can be part of a macroeconomic stabilization toolkit, particularly in economies with low public debt and where public banks are financially sound and well resourced.
- Limitations and risks:
  - High public debt constrains the countercyclical ability of public banks.
  - Potential adverse outcomes include misallocation of credit, crowding out of private credit, deterioration of loan quality in vulnerable banks, increased financial sector fragility, and the materialization of contingent fiscal liabilities (for example from extensive government guarantees).
- Policy considerations:
  - The stabilization benefits of public banks should be weighed against their potential economic and fiscal costs.
  - Maintaining public banks’ financial soundness and ensuring adequate fiscal space are important preconditions for using public banks countercyclically.

*Source: WP/21/7 — Revisiting the Stabilization Role of Public Banks: Public Debt Matters, H. Elif Ture, January 2021 (Section 1 of wpiea2021007-print-pdf).*

### Section 2

### Section 2

### Cyclical behavior of private and public bank lending — core findings
- Private banks increase lending procyclically in developing economies (coefficients for “cycle” are positive and significant for developing economies in columns (1) to (12)), while keeping lending acyclical in advanced economies (coefficients for “cycle” are smaller, mostly negative, and insignificant for advanced economies in columns (1) to (12)).
- Public banks increase lending significantly less than private banks, particularly in developing economies (coefficients for “public bank * cycle” are negative in columns (1) to (4)), but averages mask heterogeneity.
- In the decade after the GFC (in columns (11) and (12)):
  - Private bank net lending growth increases by 2 (-0.4) percentage points in developing (advanced) economies in response to a 1 percentage point increase in the growth rate of GDP per capita relative to its trend.
  - Public banks’ net lending grows 1.5 (3.8) percentage points less in developing (advanced) economies with low public debt levels in this period, yielding less procyclical (or countercyclical) lending by public banks in developing (advanced) economies with low public debt levels.
  - When public debt is high, public bank net lending grows 2.5 (2.7) percentage points more in developing (advanced) economies compared to when public debt is low; the difference is statistically significant only for developing economies.

### Heterogeneity by public debt and crisis period
- Lending by public banks is less procyclical than private bank lending in economies with low public debt levels (coefficients for “public bank * cycle” are negative in columns (5) to (12)).
- This is not the case in economies with high public debt, particularly in developing economies and outside of the GFC period (coefficients for “public bank * cycle * high public debt” are positive in columns (7) to (12)).
- Differentiating booms and busts (Table 8):
  - Private bank lending is procyclical during both booms and busts in developing economies (coefficients for “cycle+/-” are mostly positive and significant for developing economies in columns (1) to (12)).
  - In advanced economies, private bank lending is procyclical only during booms (coefficients for “cycle+” are mostly positive and significant) and countercyclical during busts (coefficients for “cycle-” are negative and significant).
  - Public bank lending is less procyclical on average, particularly during busts (coefficients for “public bank * cycle+/-” are negative in columns (1) to (4)), and also in economies with low public debt (columns (5) to (12)).
  - In economies with high public debt, public bank lending is procyclical during booms (coefficients for “public bank * cycle+ * high public debt” are mostly positive in columns (5) to (12)), and during busts in developing economies particularly after the GFC (coefficients for “public bank * cycle- * high public debt” are positive in columns (8) and (12)).

### Public commercial vs development banks
- Both public commercial and development banks exhibit similar cyclical behavior despite different business models.
- Lending by public commercial and development banks in economies with low public debt levels is less procyclical than private commercial bank lending (coefficients for “public commercial/development bank * cycle” are mostly negative in columns (5) to (12) of Table 9).
- In economies with high public debt, lending by both public commercial and development banks is not less procyclical and is particularly procyclical in developing economies and outside the GFC period (coefficients for “public commercial/development bank * cycle * high public debt” are mostly positive in columns (7) to (12) of Table 9).

### Control variables and robustness
- Control variables influence bank lending as expected:
  - Net lending grows faster in developing economies with higher financial development.
  - Faster lending is associated with lower loan size in the previous period, lower bank size (total assets), higher capitalization rate (particularly in advanced economies), lower nonperforming loan ratios, higher profitability (particularly in developing economies), higher deposit funding ratios in developing economies, and lower liquidity ratios in developing economies.
- Robustness checks:
  - Using the growth rate of net loans in constant local currency terms (Table 3) leaves main results qualitatively unchanged; coefficients for “public bank * cycle * high public debt” are positive in columns (7) to (12), particularly before the GFC. In the decade after the GFC, coefficients for developing economies (in column (12)) are no longer statistically significant.
  - Additional robustness exercises (Tables 4–7) include: (1) dropping banks without eight years of consecutive data between 2005 and 2012; (2) relaxing the assumption that development banks are public banks; (3) identifying public banks as majority (more than 50 percent) government owned; (4) using the GDP per capita growth rate as a cyclical indicator (instead of the gap). These checks do not change the main results materially.

### Interpretation, costs, and policy trade-offs
- High public debt levels could hamper the countercyclical role of public banks, particularly in developing economies and during “domestic” busts, because sovereigns and public banks face higher borrowing costs and tighter budget constraints.
- Evidence raises concerns about public banks:
  - Greater public bank ownership has been associated with lower financial development and deepening, lower economic growth, higher financial instability, and more politically motivated lending.
  - Public commercial banks operating in developing economies have lower profitability, lower interest margins, higher overhead costs, and higher nonperforming loan ratios than private banks.
  - Public banks can complicate the sovereign-bank nexus via funding government needs (often via “moral suasion”) and receiving government support, potentially creating a sovereign-bank doom loop.
- Policy implications:
  - While public banks—if financially sound and well-resourced—can help stabilize the economy, particularly in economies with low public debt, their potential countercyclical benefits must be weighed against fiscal and economic costs, including contingent liabilities and allocative inefficiencies.
  - Strengthened governance and oversight of public banks can help mitigate trade-offs.
  - To maximize effectiveness and limit fiscal risks, ensure transparency, clarify the rationale for state intervention, assess and disclose risks and costs, implement risk mitigation measures, and make budgetary provisions for materialized risks.
- Suggested direction for future research:
  - With suitable data, analyze the role of government guarantees in boosting banks’ ability to lend during crises and whether such guarantees are disproportionately utilized by public banks (including through moral suasion), potentially explaining stronger countercyclical roles during busts.

*Source: Section 2 of the provided IMF content unit (wpiea2021007-print-pdf).*

### Section 3

### wpiea2021007-print-pdf - Section 3

### Scope and data
- Sample periods reported: "1999-2018", "1999-2018 excl. GFC", "1999-2007", "2010-2018".
- Data sources: "Fitch Connect, IMF World Economic Outlook database, IMF Financial Development database, and author’s estimates."
- Public banks definition (Table notes): "Public banks are defined as those with over 25 percent of equity owned by the government." (Tables 4, 5, 7, 8)
- Alternative public bank definitions used in different tables:
  - Table 6: "Public banks are defined as those with over 50 percent of equity owned by the government."
  - Table 5: "without identifying national development banks as public banks"
- High public debt definition: "Countries with high public debt are those above the 75th percentile of the distribution across the whole sample, roughly corresponding to 100 percent of GDP for AEs and 60 percent of GDP for EMDEs."
- Country groups: "AEs = advanced economies, EMDEs = emerging market and developing economies."
- Dependent variable across tables: Loan Growth (in current US dollars, dropping banks without eight years of consecutive data between 2005 and 2012 in Table 4).

### Key estimated cyclical coefficients (selected, exact values as reported)
- Table 4 (1999-2018; AEs / EMDEs columns shown):
  - GDP pc Growth Gap (Cycle): 0.226 (AEs), -0.181 (EMDEs), 0.537*** (AEs), 0.595** (EMDEs), 0.189 (AEs), -0.185 (EMDEs), 0.511*** (AEs), 0.598** (EMDEs), 1.611***, 0.233, -0.131, 0.589 (other columns as shown).
  - Public Bank * Cycle: -0.842, -1.398***, -1.208, -0.616, -1.101*, -1.204**, -1.771**, -1.473**, -2.816***, -3.345***, -4.253**, -1.288.
  - High Public Debt (HD): 0.0192**, -0.0413, 0.0116, -0.0453*, -0.106***, -0.0740, 0.0387***, -0.0845***.
  - Public Bank * Cycle * HD: 1.319, -0.478, 3.185, 1.629*, 8.411, 4.765***, 3.208, 2.017.
  - Log of Total Assets = L: -0.172***, -0.160***, -0.166***, -0.144***, -0.175***, -0.161***, -0.168***, -0.145***, -0.324***, -0.192***, -0.209***, -0.205***.
  - Net Loans / Total Assets = L: -0.454***, -0.606***, -0.395***, -0.611***, -0.458***, -0.619***, -0.397***, -0.614***, -0.424***, -0.979***, -0.679***, -0.909***.
  - Nonperforming / Gross Loans = L: -0.517***, -0.580***, -0.504***, -0.504***, -0.519***, -0.568***, -0.503***, -0.501***, -0.744***, -0.541**, -0.518***, -0.371.
  - Constant: 0.751***, 0.765***, 0.740***, 0.570***, 0.755***, 0.782***, 0.737***, 0.585***, 1.784***, 0.737***, 0.876***, 0.781***.
  - Observations: 13,913; 3,411; 12,299; 2,973; 13,913; 3,411; 12,299; 2,973; 5,225; 817; 7,074; 2,156 (by column).
  - R-squared: 0.145, 0.164, 0.143, 0.157, 0.146, 0.165, 0.143, 0.160, 0.180, 0.166, 0.174, 0.149.
  - Number of fid: 1,931; 625; 1,890; 611; 1,931; 625; 1,890; 611; 1,101; 235; 1,485; 562.

- Table 5 (without identifying national development banks as public banks; selected):
  - GDP pc Growth Gap (Cycle): -0.0403, 0.860***, 0.456*, 1.427***, -0.310, 0.864***, 0.142, 1.437***, 0.844, 0.413, -0.673**, 1.617***.
  - Public Bank * Cycle: -0.253, -0.987***, -1.237, -0.568, -0.694, -1.227***, -1.821*, -2.187***, -5.479***, -3.413***, -2.248, -2.115**.
  - High Public Debt (HD): 0.0860***, -0.00469, 0.0705***, 0.00232, -0.161**, -0.0506, 0.129***, -0.0192.
  - Public Bank * Cycle * HD: 2.080*, 0.693, 2.665, 3.170***, 8.605**, 6.713***, 1.207, 3.935***.
  - Log of Total Assets = L: -0.172***, -0.178***, -0.157***, -0.157***, -0.185***, -0.178***, -0.166***, -0.162***, -0.304***, -0.0909, -0.244***, -0.294***.
  - Net Loans / Total Assets = L: -0.436***, -0.770***, -0.373***, -0.774***, -0.458***, -0.768***, -0.395***, -0.770***, -0.423***, -1.108***, -0.670***, -1.135***.
  - Nonperforming / Gross Loans = L: -0.690***, -0.442***, -0.762***, -0.363**, -0.712***, -0.446***, -0.765***, -0.387**, -1.029***, 0.172, -0.596***, -0.333.
  - Observations: 9,013; 1,986; 7,944; 1,715; 9,013; 1,986; 7,944; 1,715; 2,664; 419; 5,280; 1,296.
  - R-squared: 0.175, 0.298, 0.174, 0.324, 0.186, 0.298, 0.181, 0.330, 0.193, 0.226, 0.194, 0.333.

- Table 6 (using majority government ownership share to identify public banks; selected):
  - GDP pc Growth Gap (Cycle): -0.171, 1.091***, 0.208, 1.687***, -0.302, 1.087***, 0.0466, 1.699***, -0.104, 0.965**, -0.486**, 1.969***.
  - Public Bank * Cycle: -0.455, -1.018***, -0.930, -0.538, -0.451, -0.893**, -0.981, -1.527**, -2.806*, -3.279**, -1.999, -1.440*.
  - High Public Debt (HD): 0.0751***, -0.0531*, 0.0608***, -0.0458, -0.181***, -0.126, 0.122***, -0.0612*.
  - Public Bank * Cycle * HD: -0.500, -0.280, 0.0661, 1.747**, 0.514, 4.896***, 0.969, 2.434**.
  - Log of Total Assets = L: -0.200***, -0.189***, -0.188***, -0.174***, -0.209***, -0.191***, -0.194***, -0.175***, -0.359***, -0.236***, -0.261***, -0.251***.
  - Net Loans / Total Assets = L: -0.428***, -0.720***, -0.348***, -0.717***, -0.446***, -0.734***, -0.366***, -0.717***, -0.338***, -0.947***, -0.662***, -0.992***.
  - Nonperforming / Gross Loans = L: -0.660***, -0.525***, -0.699***, -0.449***, -0.666***, -0.512***, -0.692***, -0.449***, -1.177***, -0.330, -0.375***, -0.291.
  - Observations: 13,908; 3,401; 12,294; 2,965; 13,908; 3,401; 12,294; 2,965; 5,228; 813; 7,066; 2,152.
  - R-squared: 0.175, 0.284, 0.170, 0.301, 0.182, 0.285, 0.174, 0.303, 0.184, 0.208, 0.198, 0.294.

- Table 7 (using GDP per capita growth rate instead of gap as cyclical indicator; selected):
  - GDP pc Growth Gap (Cycle) coefficients reported in multiple columns: -0.166, 1.124***, 0.240, 1.729***, -0.302, 1.121***, 0.0789, 1.740***, -0.113, 1.029**, -0.425**, 1.976***.
  - Public Bank * Cycle: -0.590, -1.163***, -1.610**, -0.788*, -0.787, -1.134***, -2.048**, -1.990***, -3.139**, -4.998***, -4.091**, -1.547*.
  - Public Bank * Cycle * HD: 1.012, -0.0558, 2.902, 2.184***, 7.715, 6.660***, 2.757, 2.583**.
  - Log of Total Assets = L: -0.200***, -0.189***, -0.188***, -0.174***, -0.209***, -0.191***, -0.195***, -0.177***, -0.359***, -0.250***, -0.262***, -0.250***.
  - Observations: 13,908; 3,401; 12,294; 2,965; 13,908; 3,401; 12,294; 2,965; 5,228; 813; 7,066; 2,152.
  - R-squared: 0.176, 0.285, 0.170, 0.302, 0.182, 0.286, 0.175, 0.305, 0.185, 0.218, 0.200, 0.295.

- Table 8 (differentiating between booms and busts; selected aggregation shown across columns):
  - GDP pc Growth Gap (Cycle) and Public Bank * Cycle reported with similar patterns to prior specifications (see Table 7 replication across columns).
  - Sample and control variable coefficient patterns are consistent with previous tables' reporting style and significance markers.

### Robust patterns and controls (preserved exact reported coefficients and significance markers)
- Across multiple specifications and sample splits:
  - The interaction Public Bank * Cycle frequently has negative point estimates and many statistically significant coefficients (examples: -1.398***; -1.163***; -1.018***; -1.608**, -3.345***, -4.253**, -4.091**).
  - Log of Total Assets = L is consistently negative and statistically significant across columns (examples: -0.172***; -0.200***; -0.195***).
  - Net Loans / Total Assets = L is consistently negative and statistically significant (examples: -0.454***; -0.428***; -0.423***).
  - Nonperforming / Gross Loans = L is consistently negative and statistically significant in many columns (examples: -0.517***; -0.690***; -0.660***).
  - High Public Debt (HD) coefficients vary in sign and significance across specifications (examples: 0.0192**; 0.0751***; -0.0531*; -0.0413).

### Model features and statistical notes
- Fixed effects:
  - Table 4 reports robust standard errors in parentheses; significance markers indicated: "*** p<0.01, ** p<0.05, * p<0.1".
  - Tables 5, 6, 7, 8 include "Bank and Year Fixed Effects" = YES in reported specifications.
- Number of banks / panels reported for each specification (examples):
  - Table 4: Number of fid: 1,931; 625; 1,890; 611; etc.
  - Table 5: Number of Banks: 997; 264; 988; 261; 997; 264; 988; 261; 507; 117; 956; 257.
  - Table 6: Number of Banks: 1,930; 623; 1,889; 610; 1,930; 623; 1,889; 610; 1,102; 234; 1,484; 562.

*Source: Fitch Connect, IMF World Economic Outlook database, IMF Financial Development database, and author’s estimates.*

### Section 4

### wpiea2021007-print-pdf - Section 4 (Table 9: Cyclical Behavior of Bank Lending)

### Key definitions and sample
- Public banks are defined as those with over 25 percent of equity owned by the government.
- Countries with high public debt are those above the 75th percentile of the distribution across the whole sample, roughly corresponding to 100 percent of GDP for AEs and 60 percent of GDP for EMDEs.
- AEs = advanced economies, EMDEs = emerging market and developing economies.
- Sample periods reported: 1999-2018, 1999-2018 excl. GFC, 1999-2007, 2010-2018.
- Observations reported in tables: 13,908; 3,401; 12,294; 2,965; 5,228; 813; 7,066; 2,152 (varies by column).
- Number of banks reported: 1,930; 623; 1,889; 610; 1,930; 623; 1,889; 610; 1,102; 234; 1,484; 562 (varies by column).
- Robust standard errors in parentheses. Significance: *** p<0.01, ** p<0.05, * p<0.1.
- Bank and Year Fixed Effects = YES in all reported specifications.
- R-squared values vary by column (examples): 0.179, 0.284, 0.172, 0.302, 0.186, 0.285, 0.177, 0.306, 0.190, 0.235, 0.202, 0.298.

### Main regression findings (selected coefficients reported exactly as in source)
- Positive GDP pc Growth Gap (Cycle+)
  - Columns (various specifications): 1.226*** (0.418); 1.149*** (0.304); 1.174*** (0.395); 1.473*** (0.351); 1.135*** (0.394); 1.164*** (0.304); 1.095*** (0.374); 1.515*** (0.348); 1.706** (0.683); 1.826*** (0.671); 0.417 (0.327); 1.300*** (0.418).

- Negative GDP pc Growth Gap (Cycle-)
  - Columns (various specifications): -1.217*** (0.261); 1.090*** (0.334); -0.742** (0.326); 1.986*** (0.500); -1.381*** (0.267); 1.077*** (0.335); -0.988*** (0.334); 1.958*** (0.503); -3.047*** (0.845); -0.683 (1.312); -1.145*** (0.338); 2.459*** (0.617).

- Public Bank interactions with cycle
  - Public Bank * Cycle+: -1.007 (1.018); -1.020** (0.463); -0.413 (1.132); -0.156 (0.541); -1.556 (0.979); -0.795 (0.718); -0.841 (1.031); -0.532 (0.825); -3.663* (1.925); -1.269 (1.214); -1.783 (1.940); -0.466 (1.186).
  - Public Bank * Cycle-: -0.441 (0.609); -1.164** (0.513); -2.414** (1.098); -1.659* (0.993); -0.611 (0.652); -1.161** (0.464); -3.274** (1.421); -4.086** (1.630); -1.557 (3.228); -9.392*** (2.773); -5.025* (2.787); -2.634 (1.831).

- High Public Debt (HD)
  - Coefficients: 0.0756*** (0.00960); -0.0532* (0.0317); 0.0628*** (0.0106); -0.0388 (0.0347); -0.176*** (0.0385); -0.0836 (0.102); 0.122*** (0.0127); -0.0610* (0.0347).

- Three-way interactions involving Public Bank, Cycle, and HD (selected)
  - Public Bank * Cycle+ * HD: 5.958 (5.849); -0.363 (0.809); 5.979 (5.947); 0.924 (0.907); 15.733 (11.26); 3.149*** (1.123); 8.949* (5.039); 3.900*** (1.447).
  - Public Bank * Cycle- * HD: 0.376 (1.413); -0.118 (1.251); 2.243 (2.017); 4.111** (1.667); -3.554 (4.414); -3.674 (3.706); 1.786 (2.970); 1.937 (1.838).

- Financial development, loan growth, and bank characteristics (lagged variables)
  - Financial Development = L: -0.00809 (0.0857); 0.434*** (0.119); -0.156 (0.0983); 0.669*** (0.133); 0.0533 (0.0857); 0.443*** (0.122); -0.0506 (0.100); 0.692*** (0.133); -0.710*** (0.162); 1.261*** (0.351); 0.324** (0.156); 0.841*** (0.203).
  - Loan Growth = L: 0.0211 (0.0134); 0.0239 (0.0189); 0.0229 (0.0144); 0.0371* (0.0207); 0.0211 (0.0134); 0.0246 (0.0190); 0.0233 (0.0143); 0.0369* (0.0207); 0.0402* (0.0210); 0.0581 (0.0422); -0.0329* (0.0188); 0.00173 (0.0307).
  - Log of Total Assets = L: -0.198*** (0.0115); -0.189*** (0.0207); -0.188*** (0.0125); -0.175*** (0.0213); -0.207*** (0.0115); -0.190*** (0.0208); -0.194*** (0.0125); -0.179*** (0.0213); -0.355*** (0.0270); -0.284*** (0.0701); -0.263*** (0.0208); -0.251*** (0.0423).
  - Equity/Total Assets = L: 0.410** (0.164); -0.220 (0.222); 0.271 (0.190); -0.0195 (0.265); 0.350** (0.165); -0.213 (0.226); 0.233 (0.191); -0.00267 (0.269); 0.382 (0.297); -0.0513 (0.622); 0.538* (0.286); 0.552 (0.382).
  - Liquid Assets/Total Assets = L: 0.0791 (0.0564); -0.227*** (0.0720); 0.145** (0.0618); -0.258*** (0.0782); 0.0689 (0.0549); -0.233*** (0.0723); 0.132** (0.0607); -0.258*** (0.0788); 0.382*** (0.117); -0.335* (0.187); -0.00268 (0.0977); -0.335*** (0.104).
  - Return on Average Assets = L: 0.483 (0.368); 1.435*** (0.487); 0.580 (0.372); 1.607*** (0.516); 0.401 (0.364); 1.414*** (0.493); 0.549 (0.371); 1.434*** (0.526); 0.847 (0.883); 0.665 (0.629); -0.279 (0.362); 1.876** (0.759).
  - Deposits/Total Liabilities = L: -0.0152 (0.0325); 0.108** (0.0512); -0.0518 (0.0373); 0.130** (0.0531); -0.0565* (0.0338); 0.111** (0.0517); -0.0811** (0.0386); 0.129** (0.0530); -0.301*** (0.100); 0.205 (0.161); -0.0304 (0.0612); 0.0279 (0.0711).
  - Net Loans/Total Assets = L: -0.421*** (0.0522); -0.722*** (0.0836); -0.346*** (0.0548); -0.720*** (0.0868); -0.438*** (0.0507); -0.736*** (0.0842); -0.363*** (0.0537); -0.727*** (0.0873); -0.323*** (0.0972); -0.961*** (0.261); -0.683*** (0.0987); -1.000*** (0.128).
  - Nonperforming/Gross Loans = L: -0.666*** (0.0898); -0.523*** (0.122); -0.697*** (0.0933); -0.445*** (0.147); -0.672*** (0.0912); -0.510*** (0.123); -0.689*** (0.0951); -0.445*** (0.148); -1.076*** (0.227); -0.270 (0.272); -0.348** (0.137); -0.283 (0.220).

### Results differentiating public development and public commercial banks (alternative specification)
- GDP pc Growth Gap (Cycle) coefficients (selected): -0.163 (0.202); 1.111*** (0.198); 0.233 (0.205); 1.718*** (0.286); -0.292 (0.196); 1.109*** (0.196); 0.0712 (0.197); 1.725*** (0.284); -0.126 (0.436); 1.027** (0.464); -0.419** (0.213); 1.976*** (0.384).

- Public Development Bank * Cycle: -0.875 (0.544); -1.987*** (0.427); -1.866** (0.847); -0.896 (0.613); -1.414*** (0.452); -1.650*** (0.336); -2.375*** (0.789); -1.791** (0.720); -2.398 (1.875); -3.350* (1.815); -4.279** (2.059); -1.626* (0.947).

- Public Commercial Bank * Cycle: -0.0631 (0.973); 1 (0.340); -0.729** (1.428); -0.168 (0.472); -0.637 (1.099); 0.314 (0.442); -0.697 (2.047); 0.0450 (0.754); -1.660** (2.026); -3.870* (1.518); -4.140*** (3.421); -2.653 (1.025).

- Public Development Bank * Cycle * HD: 4.300*** (0.655); -1.192 (1.311); 5.495 (5.050); 2.513*** (0.838); 15.55*** (2.023); 2.369 (2.053); 1.526 (2.051); 2.546** (1.123).

- Public Commercial Bank * Cycle * HD: -2.335 (1.733); -0.0523 (0.592); -0.649 (2.319); 1.752* (0.907); 1.589 (2.219); 6.039*** (1.742); 2.269 (3.683); 2.512* (1.283).

- Other lagged controls and their coefficients are reported in the table (see Financial Development = L, Loan Growth = L, Log of Total Assets = L, Equity/Total Assets = L, Liquid Assets/Total Assets = L, Return on Average Assets = L, Deposits/Total Liabilities = L, Net Loans/Total Assets = L, Nonperforming/Gross Loans = L) with coefficients identical to those listed under the main specification above.

### Interpretation highlights implicit in reported coefficients
- Positive GDP per capita growth gaps (Cycle+) are generally associated with positive and often statistically significant coefficients in many specifications (e.g., 1.226***, 1.149***, 1.174***).
- Negative GDP per capita growth gaps (Cycle-) show negative and often statistically significant coefficients in many specifications (e.g., -1.217***, -1.381***), but also display heterogeneous signs and significance across specifications and subsamples.
- Interactions indicate heterogeneous cyclical lending behavior for public banks versus others, with sizable and in some cases significant three-way interactions involving High Public Debt (HD) and public bank types (examples: Public Bank * Cycle+ * HD coefficients such as 3.149***, 3.900***; Public Development Bank * Cycle * HD coefficients such as 4.300***, 2.513***, 15.55***).
- Bank-level controls (size, capitalization, liquidity, asset quality, profitability) have economically and statistically meaningful associations with lending outcomes; for example, Log of Total Assets = L is consistently negative and significant (e.g., -0.198***), and Net Loans/Total Assets = L is consistently negative and significant (e.g., -0.421***).

*Source: Table 9, "Cyclical Behavior of Bank Lending (in current US dollars) (differentiating between public development and commercial banks)", Fitch Connect, IMF World Economic Outlook database, IMF Financial Development database, and author’s estimates.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021007-print-pdf.pdf_