## wpiea2020275-print-pdf

## Source details

**Canonical URL:** [wpiea2020275-print-pdf](https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020275-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2020/english/wpiea2020275-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2020/english/wpiea2020275-print-pdf.pdf.json)

---

### I. Introduction — research question and motivation
- Research focus: implications of government interventions in banks on banks’ market power measured by the Lerner index using cross-country evidence from the Global Financial Crisis experience.
- Motivation: COVID-19–related deterioration could produce capital shortfalls and renewed government interventions; understanding effects on market power matters for credit costs, availability, and monetary policy transmission.
- Conceptual channels (net effect empirical):
  - Increase market power: interventions reduce marginal cost of funding; signal greater safety allowing higher loan rates; raise probability of future bailouts (lower market discipline).
  - Decrease market power: intervened banks compete more aggressively; moral suasion to extend lower-price services; recognition of losses under more stringent oversight.
  - Empirical question: which channel dominates?

### II. Data — scope and sample statistics
- Intervention dataset and coverage:
  - Source: Igan et al. (2019), hand-collected bank interventions, period 2007–2017.
  - Original country coverage: 37 countries; matched sample covers 27 countries after combining with bank financials.
  - Combined sample: 25,998 banks; 813 banks experienced at least one intervention.
  - Total interventions in matched sample: 1,123 interventions, by type:
    - equity injections: 773 cases
    - debt support: 43 cases
    - hybrid securities: 88 cases
    - guarantees: 83 cases
    - impaired asset relief: 272 cases
- Monetary aggregates:
  - Cumulative direct interventions amounted to $1 trillion.
  - Guarantees extended amounted to $1.5 trillion.
- Cross-country and size statistics:
  - Median number of interventions across countries: 7 (average 42).
  - Median intervention aggregated at the country level (size relative to GDP): 5 percent of GDP.
  - Largest intervention: Greece, 45 percent of GDP.
  - Smallest intervention: Lithuania, 0.1 percent of GDP.
  - Median length of interventions (time from first intervention until government stake was fully divested): 3 years.
  - Longest intervention: 11 years.
  - Shortest intervention: 1 year.
- Matching coverage and pre-intervention differences:
  - Matching coverage: 76.4 percent of intervened banks traced in the Fitch Connect dataset.
  - Pre-intervention: intervened banks tend to be larger (log of total assets), less profitable (ROA), hold more loans as a percentage of assets, and have less market power (Lerner index).

### III. Empirical methodology
- Outcome measure — Lerner index:
  - Lerner = (p − mc) / p = 1 − mc/p.
  - p defined as ratio of total income to quantity Q (assets).
  - Marginal cost mc estimated as mc_{b,t} = ε_{b,t} C_{b,t} / Q_{b,t}, where ε_{b,t} is elasticity of costs to quantity estimated from a trans-log cost function.
  - Trans-log cost function includes input price vector w_j and loan impairment charges (as share of total assets) included in w_j.
  - Controls X_{b,t}: equity over assets, loans to assets, NPLs over gross loans; bank fixed effects μ_b and year fixed effects π_t.
  - Country-specific OLS estimates used to compute ε_{b,t}.
  - Sample average Lerner index: 0.236.
  - Cross-country Lerner range example: Austria 0.027; Finland 0.452.
  - United States example: average Lerner 0.236 with standard deviation 0.112.
- Identification — matching + difference-in-differences:
  - Multivariate matching (Mahalanobis Distance) on pre-intervention Lerner, bank assets, total costs, revenues, NPLs; country fixed effects included.
  - Matching diagnostics: Kolmogorov-Smirnov and paired t-tests indicate successful balance (p-values > 0.10).
  - Difference-in-differences applied to matched sample to estimate post-intervention differential evolution.
- Event-study specification:
  - Balanced panel with 4 periods pre- and post-intervention baseline.
  - Coefficients β_j trace differences between intervened and matched banks.
  - Standard errors clustered at the bank level.
  - Identification assumption: after matching, no unobservable characteristic affects change in market power differentially; absence of differential pre-trends supported.

### IV. Main results on market power (Lerner index)
- Event-window and diff-in-diff estimates:
  - Estimated reduction in the Lerner index in the year of intervention: 0.02.
  - Intervention x Post coefficient (Table 4, Lerner): -0.020*** (0.003).
  - Reduction is statistically significant and persists in the three years post intervention.
  - Magnitude: equal to about 20 percent of the median Lerner index in the matched sample.
  - No pre-intervention difference in Lerner between treated and matched control groups.
- Decomposition: costs versus prices
  - Price component: Intervention x Post coefficient (Table 4, Price): 0.000 (0.001) — no effect on prices.
  - Marginal cost: Intervention x Post coefficient (Table 4, MC): 0.004*** (0.001) — increase in marginal cost drives Lerner decline.
  - Interpretation: reduction in ability to price over marginal costs due to higher marginal costs absorbed by the bank.

### V. Decomposition of costs, revenues, and lending
- Cost components (Table 5):
  - MC: 0.002*** (0.001)
  - Interest: 0.000 (0.001)
  - Personnel: -0.013 (0.013)
  - Other Operating: 0.0005* (0.0003)
  - Loan Impairment: 0.214*** (0.028)
  - Total Operating: 0.003*** (0.001)
- Key patterns:
  - Rise in costs is not driven by higher cost of funding nor higher wages.
  - Rise in costs is mostly driven by loan impairment charges (Table 5, Loan Impairment: 0.214*** (0.028)).
  - Dynamic: sharp increase in loan impairment charges in the year of intervention, further acceleration the year after, then gradual decline; operating expenses and loan impairment charges remain significantly higher than pre-intervention even after three years.
- Prices split (Table 6):
  - Price: 0.000 (0.001)
  - Interest: 0.001 (0.003)
  - Non-Interest: 0.000 (0.002)
  - No significant effect on interest or non-interest revenue components.
- Lending and asset quality (Table 7):
  - NPL / Loans: 0.441*** (0.112)
  - Net Loans: 0.080 (0.066)
  - Loans / Assets: 0.005 (0.007)
  - No evidence of expansion of loan portfolio; higher impairment charges associated with an increase in NPLs.
  - Given no increase in prices and no expansion in lending, change in loan impairment charges not explained by riskier lending growth.

### VI. Proposed mechanism: recognition of losses
- Conceptual mechanism:
  - Government intervention introduces a new stakeholder with incentives/mandate to restore bank health and buffers to absorb losses.
  - Banks re-evaluate loans on a lifetime expected loss basis and recognize losses.
- Empirical implication and evidence:
  - Increase in loan impairment charges without higher prices suggests banks recognize losses but do not pass higher marginal costs to borrowers — consistent with reduced market power post-intervention.

### VII. Robustness checks
- Variations and results:
  - Event window varied between 1 and 5 years: Appendix Table 2 reports Intervention x Post coefficients across windows 5, 4, 3, 2, 1 as 0.017*** (0.002), 0.020*** (0.003), -0.018*** (0.004), 0.016*** (0.003), 0.014*** (0.003) respectively (Window indicates length of window).
  - Matches per intervened bank varied between 1 and 4: Appendix Table 2 reports Intervention x Post coefficients of 0.019*** (0.002), 0.020*** (0.003), 0.020*** (0.003), 0.019*** (0.003) for # of Matches 2, 3, 2, 1 respectively.
  - Bank fixed effects (Appendix Table 3): Intervention x Post Column (1): 0.020*** (0.003); Column (2): 0.018*** (0.002).
  - Weighting each country equally (Appendix Table 4): Lerner -0.020*** (0.001); MC 0.003*** (0.0002); Price -0.0002 (0.0002); Loan Impairment 0.243*** (0.009).
  - Timing within year (Appendix Table 5): Intervention x Post x Month -0.001 (0.004); Intervention x Post x 1{Month <= 6} 0.003 (0.005) — neither significant.
  - Generated regressor and bootstrapping: 500 bootstrap samples; results remain significant at 1 percent (standard error = 0.005).
- Cross-country split (Table 8):
  - United States (Panel A, # of Treatment Obs. 642):
    - Lerner -0.018*** (0.005)
    - MC 0.003*** (0.001)
    - Price 0.000 (0.001)
    - Loan Impairment 0.228*** (0.030)
  - Non-United States (Panel B, # of Treatment Obs. 132):
    - Lerner -0.012* (0.007)
    - MC 0.005*** (0.0020)
    - Price -0.001 (0.0020)
    - Loan Impairment 0.226*** (0.0075)
  - Qualitatively similar results; larger drop in Lerner among US banks.

### VIII. Heterogeneity analyses
- By intervention type (Table 9):
  - Equity interventions (# of Treatment Obs. 650):
    - Lerner -0.019*** (0.004)
    - MC 0.003*** (0.001)
    - Price 0.004 (0.001)
  - Non-Equity interventions (# of Treatment Obs. 50):
    - Lerner -0.010 (0.013)
    - MC -0.001 (0.004)
    - Price -0.003 (0.003)
  - Key result: market power declines significantly only for banks that experienced equity injections.
- By intervention characteristics (Table 10):
  - Intervention x Post x Size coefficients:
    - Lerner: -0.663*** (0.112)
    - Loan Imp.: 2.518*** (0.905)
  - Intervention x Post x Duration coefficients:
    - Lerner: -0.005*** (0.001)
    - Loan Imp.: 0.051*** (0.010)
  - M&A interactions not significant for Lerner or Loan Impairment.
  - Larger and longer interventions produce larger effects on market power and loan impairment.
- By bank characteristics (Appendix Table 6):
  - Interactions with Market Share, Assets, ROA, Leverage Ratio, NPLs/Loans show no reported coefficient reaching conventional significance.
- By country characteristics (Table 11):
  - Intervention x Post x % change in GDP per capita during GFC: 0.850** (0.300) (measured as change between 2008 and 2009).
  - Intervention x Post x pre-GFC Leverage Ratio: -0.038 (0.129) — not significant.
  - Interpretation: decline in Lerner is larger for banks headquartered in countries with deeper economic contraction during the GFC.

### IX. Summary statistics for matched sample (Table 3)
- Matched sample (bank-year level):
  - N 18024; # of Banks 2253.
  - Lerner: 25th 0.0094, Median 0.1089, 75th 0.1900.
  - Net Loans: 25th 125.0, Median 498.0, 75th 1992.7.
  - Total Assets: 25th 178.9, Median 721.5, 75th 2899.8.
  - Deposits: 25th 144.3, Median 556.2, 75th 2026.2.
  - Net Interest Margin: 25th 3.10, Median 3.65, 75th 4.20.
  - Loan Impairment Charges / Assets: 25th 0.1212, Median 0.2989, 75th 0.7103.
  - ROE: 25th 1.680, Median 6.460, 75th 10.981.
  - ROA: 25th 0.1257, Median 0.5558, 75th 0.9521.
  - Revenue / Assets: 25th 0.0508, Median 0.0590, 75th 0.0678.
  - Marginal Cost: 25th 0.0448, Median 0.0541, 75th 0.0638.

### X. Conclusions and policy implications
- Main empirical conclusion:
  - Banks that received direct equity injections experienced a significant decline in pricing over marginal costs (Lerner decline ~0.02).
  - Intervened banks exhibited significant increases in costs due to greater recognition of loan losses (loan impairment increases ~0.214–0.243 depending on specification), not matched by higher prices.
  - Effect stronger for larger and longer interventions and in countries with sharper output losses; results robust to multiple checks.
- Policy takeaways:
  - Findings mitigate concern that interventions will automatically lead to higher lenders’ market power and associated welfare losses from higher prices or lower funding costs.
  - However, intervened banks not matching increased marginal costs with higher prices may raise concerns about long-run viability of those banks and potential financial stability implications.
  - Future research needed on long-run viability and stability implications.

*Source: wpiea2020275-print-pdf*

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

### References

### I. Introduction — research question and motivation
- Research focus: implications of government interventions in banks on banks’ market power (measured by the Lerner index) using cross-country evidence from the Global Financial Crisis experience.
- Motivation: COVID-19–related deterioration could produce capital shortfalls and renewed government interventions; understanding effects on market power matters for credit costs, availability, and monetary policy transmission.
- Conceptual channels through which interventions could affect market power:
  - Increase market power if interventions reduce marginal cost of funding, signal greater safety (allowing higher loan rates), or raise probability of future bailouts (lower market discipline and higher loan rates).
  - Decrease market power if intervened banks compete more aggressively by cutting prices, face moral suasion to extend lower-price services, or recognize losses (raising costs) under more stringent oversight.
- Net effect is empirical.

### Key empirical findings (summary)
- Market power for intervened banks decreases following government interventions.
- The decrease is more pronounced for larger and longer interventions.
- The drop in market power appears driven by a rise in costs and is not associated with any increase in prices for interest or non-interest products.
- The increase in costs is mainly the result of higher loan impairment charges.
- No observed rise in risky lending following interventions.
- Results are robust to a number of checks and are not driven by specific countries.
- Interpretation: loan impairment charges might have risen because capital injections allowed banks to recognize losses they hadn’t before or because of more stringent post-intervention oversight.

### II. Data — scope and sample statistics
- Intervention dataset: Igan et al. (2019), hand-collected bank interventions, period 2007–2017, covering 37 countries originally; matched sample covers 27 countries after combining with bank financials.
- Combined sample: 25,998 banks, 813 banks experienced at least one intervention.
- Total interventions in matched sample: 1,123 interventions, by type:
  - equity injections: 773 cases
  - debt support: 43 cases
  - hybrid securities: 88 cases
  - guarantees: 83 cases
  - impaired asset relief: 272 cases
- Monetary aggregates:
  - Cumulative direct interventions amounted to $1 trillion.
  - Guarantees extended amounted to $1.5 trillion.
- Cross-country and size statistics:
  - Median number of interventions across countries: 7 (average 42).
  - Median intervention aggregated at the country level (size relative to GDP): 5 percent of GDP.
  - Largest intervention: Greece, 45 percent of GDP.
  - Smallest intervention: Lithuania, 0.1 percent of GDP.
  - Median length of interventions (time from first intervention until government stake was fully divested): 3 years.
  - Longest intervention: 11 years.
  - Shortest intervention: 1 year.
- Matching coverage: 76.4 percent of intervened banks traced in the Fitch Connect dataset.
- Pre-intervention differences: intervened banks tend to be larger (log of total assets), less profitable (ROA), hold more loans as a percentage of assets, and have less market power (Lerner index).

### III. Empirical methodology
- Outcome measure — Lerner index:
  - Definition: Lerner = (p − mc) / p = 1 − mc/p.
  - p defined as ratio of total income to quantity Q (assets).
  - Marginal cost mc estimated as derivative of cost with respect to quantity: mc_{b,t} = ε_{b,t} C_{b,t} / Q_{b,t}, where ε_{b,t} is elasticity of costs to quantity estimated from a trans-log cost function.
  - Trans-log cost function estimated with input price vector w_j (total interest expenses over deposits; personnel expenses over assets; other operating expenses over assets) and loan impairment charges (as share of total assets) included in w_j to capture price of credit risk.
  - Cost function specification includes log Q, (log Q)^2, interactions with input prices, controls X_{b,t} (equity over assets, loans to assets, NPLs over gross loans), bank fixed effects μ_b and year fixed effects π_t.
  - Country-specific OLS estimates used to compute ε_{b,t} and hence Lerner_{b,t} = 1 − ε_{b,t} C_{b,t} / Income_{b,t}.
  - Sample average Lerner index: 0.236.
  - Cross-country Lerner range example: Austria 0.027; Finland 0.452.
  - United States example: average Lerner 0.236 with standard deviation 0.112.
- Identification strategy — matching + difference-in-differences:
  - Multivariate matching algorithm (Mahalanobis Distance) based on pre-intervention values of Lerner index, bank assets, total costs, revenues, and NPLs; country fixed effects included in matching.
  - Matching diagnostics: Kolmogorov-Smirnov and paired t-tests indicate successful balance (p-values > 0.10 across matched variables).
  - After matching, no statistical differences across observables in the pre-intervention period; no differential pre-trends (illustrated by Figures 1–3).
  - Difference-in-differences applied to matched sample to estimate post-intervention differential evolution of Lerner index and its components.

### IV. Related literature positioning
- Contrasting evidence:
  - Berger and Roman (2015): TARP-recipient banks increased market shares and market power in U.S. bank-level diff-in-diff analysis.
  - Calderon and Schaeck (2016): across many countries, government interventions associated with a decline in the Lerner index and net interest margins.
- Contribution: bank-level analysis across a large sample of countries that differentiates intervention types, sizes, and durations.

*Source: wpiea2020275-print-pdf — References section and corresponding introductory, data, and methodology text.*

### Section V). Table 3 presents summary statistics for the matched sample.

### Section V). Table 3 presents summary statistics for the matched sample.

### Empirical specifications and identification
- Event-study design (equation (6)):
  - Outcome: y_{b,t,c} for bank b in year t in country c.
  - Intervention indicator: Intervention_b (dummy if bank b received a direct government intervention).
  - Time relative to intervention: j with T the year of intervention.
  - Controls: α_c (country fixed effects), τ_t (time fixed effects).
  - Balanced panel with 4 periods pre- and post-intervention in the baseline; set of β_j coefficients trace differences between intervened and matched banks.
- Difference-in-differences (equation (7)):
  - Specification: y_{b,t,c} = α_c + τ_t + β * Intervention_b * 1{t>T} + ε_{b,t,c}.
  - β summarizes the differential evolution of the outcome after intervention.
- Standard errors clustered at the bank level.
- Identification assumption: after matching, no unobservable characteristic affects change in market power differentially between treated and untreated banks; absence of different pre-trends between control and treated groups mitigates concerns.

### Main results on market power (Lerner index)
- Event-window and main estimates:
  - Figure 1: difference in the Lerner index between intervened and matched banks over an 8-year window around the intervention.
  - Estimated reduction in the Lerner index in the year of intervention: 0.02.
  - The 0.02 reduction is statistically significant and persists in the three years post intervention.
  - Magnitude: equal to about 20 percent of the median Lerner index in the matched sample.
  - Pre-intervention: no difference in the Lerner between treated and matched control groups (difference close to zero).
- Mechanism: costs versus prices
  - The reduction in the Lerner index is entirely driven by an increase in the marginal cost of assets.
  - Prices are not affected by government interventions (see Table 4, columns 2 and 3); the cost increase is absorbed by the bank and not passed on to borrowers or other clients.
  - Interpretation: intervened banks experience a decrease in their ability to price over marginal costs.

### Decomposition of costs and revenues
- Cost components analyzed (Table 5):
  - Total costs decomposed into interest expenses and total operating costs.
  - Total operating costs further decomposed into wages, other operating expenses, and loan impairment charges.
- Findings:
  - The rise in costs is not driven by a higher cost of funding (column 2) nor by higher wages (column 3).
  - The rise in costs is mostly driven by loan impairment charges (column 5), which drive the evolution of total operating expenses (column 6).
  - Dynamic pattern: sharp increase in the year of intervention, further acceleration the year after, then gradual decline; operating expenses and loan impairment charges remain significantly higher than pre-intervention even after three years (see Figure 2).
  - Prices split into interest and non-interest revenues (Table 6): no significant effect on either component.
- Loan portfolio and asset quality (Table 7):
  - No evidence of expansion of loan portfolio: net loans and loans over assets do not increase for intervened banks.
  - Higher impairment charges are associated with an increase in NPLs.
  - Given no increase in prices, the change in loan impairment charges is not explained by an expansion or shift toward riskier loans.

### Proposed mechanism: recognition of losses
- Conceptual mechanism:
  - Government intervention introduces a new stakeholder with incentives/mandate to put the bank back in shape and buffers to absorb losses.
  - Bank re-evaluates loans on a lifetime expected loss basis and should adjust loan pricing accordingly.
- Empirical implication:
  - Evidence shows banks recognize losses (increase in impairment charges) but do not increase prices—suggesting a lack of market power post-intervention (Figure 3).

### Robustness checks
- Variations tested:
  - Event window varied between 1 and 5 years around the intervention.
  - Matches per intervened bank varied between 1 and 4.
  - Country fixed effects substituted with bank fixed effects.
- Findings:
  - Results in Appendix Tables 1 to 3 confirm baseline estimates; point estimates remain very close and significant at 1 percent.
- Cross-country dimension:
  - Sample skewed towards US institutions; sample split between US and non-US banks yields qualitatively similar results, with larger drop in Lerner among US banks (Table 8).
  - Weighting observations to give equal weights to each country yields similar conclusions (Appendix Table 4).
- Timing within year concern:
  - Augmented model with triple interaction with month of intervention; if interventions were late in the year responding to declines in Lerner, triple interaction would be significantly negative.
  - Result: point estimate for triple interaction is very close to zero (Appendix Table 5), dismissing this concern.
- Generated regressor and bootstrapping:
  - Lerner index estimated in a first stage; potential “generated regressor” issue.
  - Bootstrapping procedure: re-estimate first and second stages with 500 bootstrapped samples.
  - Reassuringly, results remain significant at 1 percent (standard error = 0.005) using bootstrap standard errors.

### Heterogeneity analyses
- Intervention characteristics:
  - Size measured by injection over assets, duration measured by years until end of public support, type split between equity and non-equity injections.
  - Asset purchases used to capture government stake more directly.
  - Key result: market power declines significantly only for banks that experienced equity injections and not for other interventions (debt or hybrid, Table 9).
  - Larger and longer interventions produce larger effects on market power and cost components (Table 10).
- Bank-level heterogeneity:
  - Tested pre-intervention bank characteristics: market share, size, profitability, leverage, asset quality.
  - Results do not vary based on pre-intervention bank characteristics (Appendix Table 6).
  - Lack of difference by leverage confirmed at country level (Table 11, column 1).
- Country-level heterogeneity:
  - Decline in Lerner is larger for banks headquartered in countries with deeper economic contraction during the GFC.
  - Quantified effect: the decline in Lerner is 0.0085 greater for an intervened bank in a country where GDP per capita fell by an additional 1 percent (Table 11, column 2).
  - Interpretation: in countries with more severe crises, larger increases in credit risk and prompt recognition of loans led to larger increases in marginal costs of lending.

### Conclusions and policy implications
- Context: COVID-19 could lead to significant bank losses; governments may intervene to support banks and limit a financial crisis-imposed amplification of the downturn.
- Main empirical conclusion:
  - Banks that received direct equity injections experienced a significant decline in pricing over marginal costs.
  - Intervened banks exhibited significant increases in costs due to greater recognition of loan losses, not matched by higher prices.
  - Effect stronger for larger and longer interventions and in countries with sharper output losses; results robust to multiple checks.
- Policy takeaways:
  - Findings mitigate concern that interventions will automatically lead to higher lenders’ market power and associated welfare losses (e.g., if intervened banks were perceived safer and charged higher prices or obtained lower funding costs).
  - However, intervened banks not matching increased marginal costs with higher prices may raise concerns about long-run viability of those banks and potential financial stability implications.
  - Future research needed on long-run viability and stability implications.

*Source: wpiea2020275-print-pdf - Section V). Table 3 presents summary statistics for the matched sample.*

### References

### References

### Selected citations
- Acharya, V., L. Borchert, M. Jager and S. Steffen (2020). Kicking the Can Down the Road: Government Interventions in the European Banking Sector, The Review of Financial Studies, forthcoming.
- Aghion, P., N. Bloom, R. Blundell, R. Griffith and P. Howitt (2005). Competition and Innovation: An Inverted-U Relationship, The Quarterly Journal of Economics, 120(2): 701–728.
- Allen, F. and D. Gale (2004). Competition and Financial Stability, Journal of Money, Credit and Banking, 36: 453–480.
- Ashraf Q. and O. Galor (2013). The 'Out of Africa' Hypothesis, Human Genetic Diversity, and Comparative Economic Development, American Economic Review, 103(1): 1–46.
- Bassett, W., S. Demiralp and N. Lloyd (2020). Government support of banks and bank lending, Journal of Banking and Finance, 112: 105–177.
- Bayazitova, D. and A. Shivdasani (2012). Assessing TARP, The Review of Financial Studies, 25(2): 377–407.
- Beck, T., O. De Jonghe and G. Schepens (2013). Bank competition and stability: cross-country heterogeneity, Journal of Financial Intermediation, 22: 218–244.
- Berger, A.N., L. Klapper, and R. Turk-Ariss (2009). Bank competition and financial stability, Journal of Financial Services Research, 35(2): 99–118.
- Berger, A.N. and R.A. Roman (2015). Did TARP Banks Get Competitive Advantages?, The Journal of Financial and Quantitative Analysis, 50(6): 1199–1236.
- Berger, A.N. and R.A. Roman (2017). Did Saving Wall Street Really Save Main Street? The Real Effects of TARP on Local Economic Conditions, The Journal of Financial and Quantitative Analysis, 52(5): 1827–1867.
- Berger, A.N., T. Makaew, and R.A. Roman (2019). Do Business Borrowers Benefit from Bank Bailouts?: The Effects of TARP on Loan Contract Terms, Financial Management, 48(2): 575–639.
- Berger, A.N., R.A. Roman and J. Sedunov (2020). Did TARP Reduce or Increase Systemic Risk? The Effect of Government Aid on Financial System Stability, Journal of Financial Intermediation, 43: 100810.
- Black, L. and L. Hazelwood (2013). The effect of TARP on bank risk-taking, Journal of Financial Stability, 9: 790–803.
- Boyd, J.H. and G. De Nicolò (2005). The Theory of Bank Risk Taking and Competition Revisited, Journal of Finance, 60(3): 1329–43.
- Brandao-Marques, L., R. Correa, and H. Sapriza (2020). Government support, regulation, and risk taking in the banking sector, Journal of Banking and Finance, 112: 105284.
- Calderon, C. and K. Schaeck (2016). The effects of government interventions in the financial sector on banking competition and the evolution of zombie banks, The Journal of Financial and Quantitative Analysis, 51: 1391–1436.
- Calomiris, C.W. and U. Khan (2015). An Assessment of TARP Assistance to Financial Institutions, Journal of Economic Perspectives, 29(2): 53–80.
- Carbo-Valverde, S., F. Rodriguez-Fernandez, and G.F. Udell (2009). Bank market power and SME financing constraints, Review of Finance 13(2): 309–340.
- Deli, Y., M. Delis, I. Hasan, and L. Liu (2019). Enforcement of banking regulation and the cost of borrowing, Journal of Banking and Finance, 101: 147–160.
- Delis, M., S. Kokas, and S. Ongena (2016). Foreign ownership and market power in banking: evidence from a world sample, Journal of Money, Credit and Banking, 48: 449–483.
- Dell’Ariccia, G., M. S. Martinez Peria, D. Igan, E. Addo Awadzi, M. C. Dobler, and D. Sandri (2018). Trade-offs in Bank Resolution, IMF Staff Discussion Note 18/02.
- Diamond, A. and J.S. Sekhon (2013). Genetic matching for estimating causal effects: A general multivariate matching method for achieving balance in observational studies, Review of Economics and Statistics, 95(3): 932–945.
- Drechsler, I., A. Savov, and P. Schnabl (2017). The Deposits Channel of Monetary Policy, Quarterly Journal of Economics, 132(4): 1819–76.
- Duchin, R. and D. Sosyura (2012). The politics of government investment, Journal of Financial Economics, 106(1): 24–48.
- Duchin, R. and D. Sosyura (2020). Safer ratios, risker portfolios: Banks’ response to government aid, Journal of Financial Economics, 113(1): 1–28.
- Gourinchas, P-O., V. Penciakova, S. Kalemli-Ozcan, and N. Sander (2020). COVID-19 and SME Failures, NBER Working Paper No. 27877.
- Gropp, R., H. Hakenes, and I. Schnabel (2010). Competition, Risk-shifting, and Public Bail-out Policies, Review of Financial Studies, 24(6): 2084–2120.
- Gross, M., D. Laliotis, M. Leika, and P. Lukyantsau (2020). Expected Credit Loss Modeling from a Top-Down Stress Testing Perspective, IMF Working Paper No. 20/111.
- Hakenes, H. and I. Schnabel (2010). Banks without parachutes: Competitive effects of government bail-out policies, Journal of Financial Stability, 6(3): 156–168.
- Igan, D., H. Moussawi, A. Tieman, A. Zdzienicka, G. Dell’Ariccia, ad P. Mauro (2019). The Long Shadow of the Global Financial Crisis: Public Interventions in the Financial Sector, IMF Working Paper No. 19/164.
- International Monetary Fund (2020). Global Financial Stability Report, October, Washington DC.
- Koettner, M. and F. Noth (2016). Did TARP Distort Competition among Sound Unsupported Banks?, Economic Inquiry, 54(2): 994–1020.
- Li, L. (2013). TARP funds distribution and bank loan supply, Journal of Banking and Finance, 37(12): 4777–92.
- Love, I. and M. S. Martinez Peria (2015). How Bank Competition Affects Firms’ Access to Finance, The World Bank Economic Review, 29(3): 413–448.
- Pagan, Adrian (1984). Econometric issues in the analysis of regressions with generated regressors, International Economic Review, 25(1): 221–247.
- Nickell, S. (1996). Competition and Corporate Performance, Journal of Political Economy, 104: 724–746.
- Wang, Y., T. Whited, Y. Wu and K. Xiao (2020). Bank Market Power and Monetary Policy Transmission: Evidence from a Structural Estimation. NBER Working Paper No. 27258.

### Figures and Tables — notes and key reported statistics
- Figure 1: plots coefficients 훽_r and standard errors from the event study specification in Section III (equation 6) with the Lerner index as y_b,t,c. Country fixed effects are included.
- Figure 2: plots coefficients 훽_r and standard errors from the event study specification in Section III (equation 6) with:
  - (a) total operating expenses over assets as y_b,t,c, where total operating expenses = sum of personnel and other operating expenses (non-interest expenses) + loan impairment charges + equity-accounted profit/loss;
  - (b) loan impairment charges over assets as y_b,t,c. Country fixed effects are included.
- Figure 3: plots coefficients 훽_r and standard errors from the event study specification in Section III (equation 6) with country market share as y_b,t,c. Country fixed effects are included.

- Table 1: Differences in pre-intervention characteristics across banks in full sample (OLS estimates at bank level; pre-intervention means over years 2004-2006)
  - Intervention coefficients: Assets 1.599*** (0.077), ROA -0.156*** (0.042), NPL/Loans -0.214** (0.086), Lerner -0.034*** (0.004).
  - Country FE: Yes. Observations: 16964. Adjusted R2: 0.513 (Assets), 0.092 (ROA), 0.516 (NPL/Loans), 0.056 (Lerner).
  - Standard errors clustered by bank. * p < 0.1, ** p < 0.05, *** p < 0.01.

- Table 2: Differences in pre-intervention characteristics across banks in matched sample (OLS estimates; mean over 4 pre-intervention periods)
  - Intervention coefficients: Assets 0.462 (0.316), ROA 0.025 (0.052), NPL/Loans -0.095 (0.111), Lerner -0.002 (0.004).
  - Country FE: Yes. Observations: 2253. Adjusted R2: 0.770 (Assets), 0.035 (ROA), 0.185 (NPL/Loans), 0.078 (Lerner).

- Table 3: Summary statistics for matched sample (bank-year level)
  - N 18024; # of Banks 2253.
  - Lerner: 25th 0.0094, Median 0.1089, 75th 0.1900.
  - Net Loans: 25th 125.0, Median 498.0, 75th 1992.7.
  - Total Assets: 25th 178.9, Median 721.5, 75th 2899.8.
  - Deposits: 25th 144.3, Median 556.2, 75th 2026.2.
  - Net Interest Margin: 25th 3.10, Median 3.65, 75th 4.20.
  - Loan Impairment Charges / Assets: 25th 0.1212, Median 0.2989, 75th 0.7103.
  - ROE: 25th 1.680, Median 6.460, 75th 10.981.
  - ROA: 25th 0.1257, Median 0.5558, 75th 0.9521.
  - Revenue / Assets: 25th 0.0508, Median 0.0590, 75th 0.0678.
  - Marginal Cost: 25th 0.0448, Median 0.0541, 75th 0.0638.

- Table 4: Difference-in-differences effect of banking interventions on Lerner index (OLS at bank-year level on matched sample)
  - Dependent Variable: Lerner (1), Price (2), MC (3).
  - Intervention x Post coefficients: Lerner -0.020*** (0.003), Price 0.000 (0.001), MC 0.004*** (0.001).
  - Country FE: Yes. Time FE: Yes. Window 4. # of Matches 2. # of Treatment Obs. 730.
  - Adjusted R2: 0.215 (Lerner), 0.559 (Price), 0.420 (MC).

- Table 5: Difference-in-differences effect on components of cost (OLS at bank-year level on matched sample)
  - Dependent Variables and Intervention x Post coefficients (standard errors):
    - MC: 0.002*** (0.001)
    - Interest: 0.000 (0.001)
    - Personnel: -0.013 (0.013)
    - Other Operating: 0.0005* (0.0003)
    - Loan Impairment: 0.214*** (0.028)
    - Total Operating: 0.003*** (0.001)
  - Country and Time FE: Yes. Window 4. # of Matches 2. # of Treatment Obs. 730.

- Table 6: Difference-in-differences effect on components of price (OLS at bank-year level on matched sample)
  - Dependent Variables and Intervention x Post coefficients:
    - Price: 0.000 (0.001)
    - Interest: 0.001 (0.003)
    - Non-Interest: 0.000 (0.002)
  - Country and Time FE: Yes. Window 4. # of Matches 2. # of Treatment Obs. 730. Adjusted R2: 0.559, 0.512, 0.498 respectively.

- Table 7: Difference-in-differences effect on lending (OLS at bank-year level on matched sample)
  - Dependent Variables and Intervention x Post coefficients:
    - NPL / Loans: 0.441*** (0.112)
    - Net Loans: 0.080 (0.066)
    - Loans / Assets: 0.005 (0.007)
  - Country and Time FE: Yes. Window 4. # of Matches 2. # of Treatment Obs. 730. Adjusted R2: 0.339, 0.540, 0.128 respectively.

- Table 8: Difference-in-differences — United States vs non-United States (matched sample)
  - Panel A: United States (# of Treatment Obs. 642)
    - Intervention x Post coefficients:
      - Lerner -0.018*** (0.005)
      - MC 0.003*** (0.001)
      - Price 0.000 (0.001)
      - Loan Impairment 0.228*** (0.030)
    - Adjusted R2: Lerner 0.200, MC 0.157, Price 0.188, Loan Impairment 0.241.
  - Panel B: Non-United States (# of Treatment Obs. 132)
    - Intervention x Post coefficients:
      - Lerner -0.012* (0.007)
      - MC 0.005*** (0.0020)
      - Price -0.001 (0.0020)
      - Loan Impairment 0.226*** (0.0075)
    - Adjusted R2: Lerner 0.520, MC 0.857, Price 0.188, Loan Impairment 0.499.

- Table 9: Heterogeneity by intervention type (matched sample; equity vs non-equity)
  - Equity interventions (Columns 1–3; # of Treatment Obs. 650):
    - Intervention x Post coefficients: Lerner -0.019*** (0.004), MC 0.003*** (0.001), Price 0.004 (0.001).
  - Non-Equity interventions (Columns 4–6; # of Treatment Obs. 50):
    - Intervention x Post coefficients: Lerner -0.010 (0.013), MC -0.001 (0.004), Price -0.003 (0.003).

- Table 10: Heterogeneity by intervention characteristics (matched sample)
  - Intervention x Post x Size coefficients:
    - Lerner: -0.663*** (0.112)
    - Loan Imp.: 2.518*** (0.905)
  - Intervention x Post x Duration coefficients:
    - Lerner: -0.005*** (0.001)
    - Loan Imp.: 0.051*** (0.010)
  - Intervention x Post x M&A coefficients:
    - Lerner: -0.001 (0.008)
    - Loan Imp.: -0.073 (0.062)
  - Country and Time FE: Yes. Window 4. # of Matches 2. # of Treatment Obs. 730.

- Table 11: Heterogeneity by country characteristics (matched sample)
  - Intervention x Post x % change in GDP per capita during GFC: 0.850** (0.300).
  - Intervention x Post x pre-GFC Leverage Ratio: -0.038 (0.129).
  - Country and Time FE: Yes. Window 4. # of Matches 2. # of Treatment Obs. 730. Adjusted R2: 0.212 (Column 1), 0.213 (Column 2).

- Appendix notes:
  - Appendix Figure 1: plots coefficients from a regression of an indicator for being an unmatched (or missing) bank in FitchConnect on ROA, capital ratio, and log assets (sample: all intervened banks from Igan et al. (2019)).
  - Appendix Figure 2: plots coefficients from a regression of an indicator for being an unmatched (or missing) bank in FitchConnect on the size of the intervention by intervention type (sample: all intervened banks from Igan et al. (2019)).
  - Appendix Table 2: Robustness for event window reports Intervention x Post coefficients: 0.017*** (0.002), 0.020*** (0.003), -0.018*** (0.004), 0.016*** (0.003), 0.014*** (0.003) across windows 5, 4, 3, 2, 1 respectively (Window indicates length of window, number of periods pre and post intervention).
  - Appendix Table 2 (robustness for number of matches): Intervention x Post coefficients: 0.019*** (0.002), 0.020*** (0.003), 0.020*** (0.003), 0.019*** (0.003) for # of Matches 2, 3, 2, 1 respectively.

*Content unit: wpiea2020275-print-pdf - References (extracted from source PDF).*

### Appendix Table 3: Robustness to bank fixed effects

### Appendix Table 3: Robustness to bank fixed effects

### Main findings
- Intervention x Post:
  - Column (1): 0.020*** (0.003)
  - Column (2): 0.018*** (0.002)
- Significance: *** p < 0.01
- Interpretation: Positive and statistically significant effect of government intervention on Lerner in both specifications.

### Specification details
- Dependent Variable: Lerner
- Country FE: Yes in Column (1); No in Column (2)
- Bank FE: No in Column (1); Yes in Column (2)
- Time FE: Yes in both columns
- Window: 4
- # of Matches: 2
- Adjusted R2:
  - Column (1): 0.215
  - Column (2): 0.620

### Notes
- OLS estimates of equation (7) at the bank-year level on the matched sample as explained in Section III.
- Intervention indicates that the bank is a recipient of a government intervention. Post indicates the year is after the bank intervention.
- Lerner is constructed as in Section IIIa.
- Standard errors, clustered by bank, in parentheses.

---

### Appendix Table 4: Robustness to weighting each country equally

### Main findings (Intervention x Post coefficients)
- Lerner (Column (1)): -0.020*** (0.001)
- MC (Marginal Cost) (Column (2)): 0.003*** (0.0002)
- Price (Column (3)): -0.0002 (0.0002)
- Loan Impairment (Column (4)): 0.243*** (0.009)
- Significance markers: Columns (1), (2), and (4) show *** p < 0.01; Column (3) not significant.

### Specification details
- Dependent Variables: Lerner, MC, Price, Loan Impairment (Columns (1)–(4))
- Country FE: Yes for all columns
- Time FE: Yes for all columns
- Window: 4 for all columns
- # of Matches: 2 for all columns
- # of Treatment Obs.: 730 for all columns
- Adjusted R2:
  - Column (1): 0.209
  - Column (2): 0.431
  - Column (3): 0.573
  - Column (4): 0.233

### Notes
- Each observation is weighted by the inverse of the number of country banks in the matched sample.
- Lerner, Price and Marginal Cost (MC) constructed as in Section IIIa.
- Standard errors, clustered by bank, in parentheses.

---

### Appendix Table 5: Robustness to timing of intervention within year

### Main findings (timing interactions)
- Intervention x Post x Month: -0.001 (0.004)
- Intervention x Post x 1{Month <= 6}: 0.003 (0.005)
- Neither timing interaction is statistically significant at conventional levels.

### Specification details
- Dependent Variable: Lerner (Columns (1) and (2))
- Country FE: Yes in both columns
- Time FE: Yes in both columns
- Window: 4
- # of Matches: 2
- # of Treatment Obs.: 730 for both columns
- Adjusted R2:
  - Column (1): 0.214
  - Column (2): 0.213

### Notes
- Month corresponds to the month number of the intervention from 1 to 12 (January is 1).
- 1{Month <= 6} is an indicator for the first half of the year.
- Standard errors, clustered by bank, in parentheses.

---

### Appendix Table 6: Heterogeneity by bank characteristics

### Main findings (Intervention x Post x characteristic)
- Market Share (Column (1)): 0.096 (0.159)
- Assets (Column (2)): 0.000 (0.001)
- ROA (Column (3)): -0.001 (0.003)
- Leverage Ratio (Column (4)): 0.0004 (0.001)
- NPLs/Loans (Column (5)): -0.0003 (0.000)
- No reported coefficient reaches conventional statistical significance based on provided standard errors.

### Specification details
- Dependent Variable: Lerner (Columns (1)–(5))
- Interaction variables sourced from FitchConnect.
- Country FE: Yes in all columns
- Time FE: Yes in all columns
- Window: 4 for all columns
- # of Matches: 2 for all columns
- # of Treatment Obs.: 730 for all columns
- Adjusted R2:
  - Column (1): 0.212
  - Column (2): 0.214
  - Column (3): 0.309
  - Column (4): 0.220
  - Column (5): 0.224

### Notes
- % change in GDP per capita during GFC is measured as the change between 2008 and 2009.
- Lerner is constructed as in Section IIIa.
- Standard errors, clustered by bank, in parentheses.

*Source: Appendix Tables 3–6 from the provided content unit.*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2020/english/wpiea2020275-print-pdf.pdf_