## 10. Liquidity Creation and Bank Failures: Effect of Bank Concentration

## Source details

**Canonical URL:** [10. Liquidity Creation and Bank Failures: Effect of Bank Concentration](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15103.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15103.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15103.pdf.json)

---

### I. Hypotheses and theoretical framing
- Three hypotheses on bank fragility:
  - Weak Fundamentals Hypothesis (WFH): failing banks show poor fundamentals (CAMELS components).
  - Liquidity Shortage Hypothesis (LSH): solvent banks fail due to depositor runs financing illiquid assets with liquid liabilities.
  - High Liquidity Creation Hypothesis (HLCH): a bank’s vulnerability increases when liquidity creation reaches high levels relative to the system.
- Theoretical and empirical context:
  - Diamond and Rajan (2000, 2001, 2002): illiquid assets funded by demand deposits increase fragility.
  - Allen and Gale (2004): intermediaries as liquidity creators are prone to failure; liquidity provision regulation can be justified.
  - Berger and Bouwman (2009, 2011): bank liquidity creation can expand before crises; high aggregate liquidity creation predicts crises better than GDP, federal funds rate, or stock returns.
- Policy framing implications:
  - WFH → prudential macroeconomic and supervisory policies.
  - LSH → confidence-building mechanisms (deposit insurance, lender-of-last-resort).
  - HLCH → heightened monitoring of liquidity creation; screen banks by liquidity creation ranking for enhanced oversight.

### II. Data and liquidity creation measurement
- Sample and scope:
  - Source: bank-level financial statement data for Russian banks from Interfax (based on CBR).
  - Original panel: quarterly from Q1 1999 to Q4 2009; analysis uses 2000–2007 to exclude 1998 crisis and 2008 global crisis effects.
  - Final sample after filters: over 33,000 bank-quarter observations.
  - Failures identified from CBR’s list of failed banks (roughly 230 failed institutions over the whole period); Russia experienced over 200 bank failures between 2000 and 2007.
- Filters applied:
  - Drop observations with total loans to total assets < 5%.
  - Exclude observations with sum of all deposits = 0.
  - Drop observations with capital-to-assets ratio > 100%.
- Liquidity creation measures (Berger and Bouwman, 2009 functional form):
  - Liquidity Creation = {½ × Illiquid Assets + 0 × Semi-Liquid Assets – ½ × Liquid Assets} + {½ × Liquid Liabilities + 0 × Semi-Liquid Liabilities – ½ × Illiquid Liabilities} – ½ × Capital
  - Two measures normalized by total assets:
    - LC1: category-based classification (liquid, semi-liquid, illiquid).
    - LC2: maturity-based classification (liabilities and interbank loans split by maturities).
- Representative classification notes:
  - Liquid assets include correspondent accounts, government securities, promissory notes.
  - Corporate loans classified as illiquid; consumer, government, interbank loans as semi-liquid in LC1.
  - LC2: interbank loans < 1 week = liquid; > 1 year and nonperforming = illiquid.
- High/low liquidity creator dummies:
  - LC_80-90% and LC_90-100% = 1 if a bank’s LC in a quarter lies in the 80th–90th percentile and above the 90th percentile, respectively.
  - LC_10-20% and LC_0-10% = 1 for 10th–20th percentile and below the 10th percentile, respectively.
- Control variables:
  - Size = logarithm of total assets.
  - ROA = return on assets.
  - Household Income Growth and Small Business Growth (regional, branch-weighted).
  - Regional Herfindahl-Hirschmann index for assets (Herfindahl) used in robustness checks.
  - Quarter and year dummies; Deposit Insurance dummy (=1 for quarters starting from beginning of 2005) used in robustness checks.
- Definition of failure:
  - Baseline: bank license withdrawn in a given quarter (dummy = 1).
  - Alternative: equity-to-assets ratio below 2% (used in robustness checks).

### III. Key empirical findings
- Aggregate liquidity creation (2000–2007):
  - LC1 rises from 22 to 30 percent of assets.
  - LC2 hovers around 18 to 21 percent of assets.
  - LC1 consistently larger than LC2 and more volatile.
- Baseline panel logit results (four quarterly lags before failure):
  - LC_90-100%:
    - Coefficient positive and significant at the 1% level across all quarters prior to failure using both LC1 and LC2.
    - Representative marginal effects (selected):
      - LC1/Assets, 1 quarter: 1.355*** [0.205]
      - LC1/Assets, 4 quarters: 1.093*** [0.225]
      - LC2/Assets, 1 quarter: 1.259*** [0.212]
      - LC2/Assets, 4 quarters: 1.122*** [0.226]
  - LC_0-10%:
    - Coefficient positive and significant in about half of baseline estimations — limited evidence that very low liquidity creation also increases failure probability.
  - Control variables:
    - Size: negative and significant in all estimations (examples: -0.255*** [0.054], -0.294*** [0.083]).
    - ROA: negative and significant in all estimations (examples: -6.009*** [0.962], -6.278*** [1.020]).
    - Household Income Growth and Small Business Growth: not statistically significant in baseline.
- Robustness checks summary:
  - Longer lags (15, 18, 21, 24 months):
    - LC_90-100% remains positive and significant across estimations (examples: LC1/24 months: 0.846*** [0.230]; LC2/15 months: 0.820*** [0.238]).
    - LC_0-10% positive and significant only at the 21-month lag.
  - Alternative failure definition (equity/total assets < 2%):
    - LC_90-100% remains positive and significant (examples: LC1/1 quarter: 1.906*** [0.340]; LC1/3 quarters: 2.056*** [0.356]).
    - LC_0-10% not consistently significant.
  - Moscow-only sample:
    - LC_90-100% positive and significant across estimations (examples: LC1/1 quarter: 1.229*** [0.248]; LC2/1 quarter: 1.305*** [0.250]).
    - Low liquidity creation has no impact.
  - Domestic private banks only:
    - LC_90-100% positive and significant at the 1% level (examples: LC1/1 quarter: 1.361*** [0.207]; LC2/1 quarter: 1.247*** [0.211]).
    - LC_0-10% positive and significant in most estimations for this subsample.
  - Deposit insurance (dummy from beginning of 2005):
    - Deposit Insurance coefficient not significant in most specifications (examples: -0.220 [0.478]; -0.433 [0.488]).
    - LC_90-100% remains positive and significant across estimations.
  - Bank concentration (regional Herfindahl):
    - LC_90-100% positive and highly significant across estimations (examples: LC1/1 quarter: 1.370*** [0.197]; LC2/1 quarter: 1.290*** [0.196]).
    - Herfindahl coefficient significant and negative (examples: LC1/1 quarter: -3.933*** [0.659]; LC2/1 quarter: -4.018*** [0.663]) — consistent with concentration reducing probability of bank failure in Russia.
- Synthesis:
  - Consistent and robust evidence supports HLCH: banks above the 90th percentile of liquidity creation face a significantly higher probability of failure under normal economic conditions.
  - Evidence linking very low liquidity creation to failures is weaker and specification-sensitive.

### IV. Policy implications and recommendations
- Regulatory monitoring:
  - Screen banks by liquidity creation ranking in the system rather than using a single absolute cutoff.
  - Place high liquidity creators (above the 90th percentile in a given quarter) on watch lists for enhanced supervision.
- Early warning systems:
  - Incorporate liquidity creation measures (LC1 and LC2 normalized by total assets) into early warning frameworks to identify banks whose core liquidity-creating activity may elevate failure risk.
- Prudential actions:
  - Consider interventions to reduce excessively high liquidity creation to limit the probability of failure and potential taxpayer costs from resolution.
  - For high liquidity creators, regulators may prefer preventive oversight or orderly wind-downs to limit systemic spillovers and preserve aggregate liquidity creation.
- Deposit insurance and safety nets:
  - Introduction of deposit insurance in Russia (beginning of 2005) did not show a statistically significant effect on failure probability in this analysis; policymakers should weigh deposit insurance’s two opposing effects (reducing depositor runs vs. encouraging risk-taking).
- Market structure:
  - Given the negative and significant relation between regional concentration (Herfindahl) and failure probability, authorities should account for local concentration dynamics when assessing systemic vulnerability.

### V. Conclusion
- The paper documents robust evidence that high liquidity creation by individual banks increases the probability of bank failure in Russia during 2000–2007, a period without major exogenous shocks to the financial system.
- The HLCH complements WFH and LSH by focusing on the core liquidity-creating activity of banks as an endogenous source of fragility.
- Policy relevance: rank-based screening of liquidity creators, integrating liquidity creation into supervision and early warning systems, and targeted oversight of high liquidity creators to prevent bank failures and limit taxpayer losses.

*Source: 10. Liquidity Creation and Bank Failures: Effect of Bank Concentration (chapter content). PDF: _wp15103 - 10. Liquidity Creation and Bank Failures: Effect of Bank Concentration.*

### REFERENCES ___________________________________________________________17

### _wp15103 - REFERENCES ___________________________________________________________17

### Tables
- 1. Liquidity Creation Measures ................................................................................................23
- 2. Development of the Main Variables ....................................................................................24
- 3. Descriptive statistics of the banks with high liquidity creation and the other 
.....................25
- 4. Liquidity Creation and Bank Failures: Baseline Model 
......................................................26
- 5. Liquidity Creation and Bank Failures: Longer Lags ...........................................................27
- 6. Liquidity Creation and Bank Failures: Alternative Measure of Failure ..............................28
- 7. Liquidity Creation and Bank Failures: Moscow Banks Only ..............................................29
- 8. Liquidity Creation and Bank Failures: Dom
estic Private Banks Only ................................30
- 9. Liquidity Creation and Bank Failures: Effect of Deposit Insurance 
....................................31

*Source: _wp15103 - REFERENCES ___________________________________________________________17*

### 10. Liquidity Creation and Bank Failures: Effect of Bank Concentration

### 10. Liquidity Creation and Bank Failures: Effect of Bank Concentration

### I. Hypotheses and theoretical framing
- Three hypotheses on bank fragility discussed:
  - Weak Fundamentals Hypothesis (WFH): failing banks show poor fundamentals (CAMELS components).
  - Liquidity Shortage Hypothesis (LSH): solvent banks fail due to depositor runs financing illiquid assets with liquid liabilities.
  - High Liquidity Creation Hypothesis (HLCH) proposed in this paper: a bank’s vulnerability increases when liquidity creation reaches high levels relative to the system.
- Theoretical support:
  - Diamond and Rajan (2000, 2001, 2002): carrying illiquid assets funded by demand deposits increases fragility.
  - Allen and Gale (2004): intermediaries as liquidity creators are prone to failure; regulation of liquidity provision can be justified.
- Empirical precedents:
  - Berger and Bouwman (2009, 2011): bank liquidity creation can expand before crises; high aggregate liquidity creation predicts crises better than GDP, federal funds rate, or stock returns.
- Policy framing:
  - WFH → prudential macroeconomic and supervisory policies.
  - LSH → confidence-building mechanisms (deposit insurance, lender-of-last-resort).
  - HLCH → heightened monitoring of liquidity creation; screen banks by liquidity creation ranking for enhanced oversight.

### II. Data and liquidity creation measurement
- Sample and scope:
  - Source: bank-level financial statement data for Russian banks from Interfax (based on CBR).
  - Original panel: quarterly from Q1 1999 to Q4 2009; analysis uses 2000–2007 to exclude 1998 crisis and 2008 global crisis effects.
  - Final sample after filters: over 33,000 bank-quarter observations.
  - Failures identified from CBR’s list of failed banks (roughly 230 failed institutions over the whole period of study); Russia experienced over 200 bank failures between 2000 and 2007.
- Filters applied:
  - Drop observations with total loans to total assets < 5%.
  - Exclude observations with sum of all deposits = 0.
  - Drop observations with capital-to-assets ratio > 100%.
- Liquidity creation measures (following Berger and Bouwman, 2009):
  - Functional form (Equation 1):
    - Liquidity Creation = {½ × Illiquid Assets + 0 × Semi-Liquid Assets – ½ × Liquid Assets} + {½ × Liquid Liabilities + 0 × Semi-Liquid Liabilities – ½ × Illiquid Liabilities} – ½ × Capital
  - Two measures:
    - LC1: category-based classification of balance sheet items (liquid, semi-liquid, illiquid).
    - LC2: maturity-based classification (liabilities and interbank loans split by maturities).
  - Both LC1 and LC2 normalized by total assets.
- Representative classification notes (selected):
  - Liquid assets include correspondent accounts, government securities, promissory notes.
  - Corporate loans classified as illiquid; consumer, government, interbank loans as semi-liquid in LC1.
  - LC2 uses maturity breaks: interbank loans < 1 week = liquid; > 1 year and nonperforming = illiquid.
- High/low liquidity creator dummies:
  - LC_80-90% and LC_90-100% = 1 if LC for a bank in a quarter lies between the 80th–90th percentile and above the 90th percentile, respectively.
  - LC_10-20% and LC_0-10% = 1 if LC in 10th–20th percentile and below the 10th percentile, respectively.
- Control variables:
  - Bank size: logarithm of total assets (Size).
  - Profitability: return on assets (ROA).
  - Local market environment: Household Income Growth (regional household income per capita) and Small Business Growth (number of SMEs × average employees), weighted by branch distribution.
  - Regional Herfindahl-Hirschmann index for assets (Herfindahl) used in robustness checks.
  - Dummy variables for quarters and years; Deposit Insurance dummy (=1 for quarters starting from beginning of 2005) used in robustness checks.
- Definition of failure:
  - Baseline: bank license withdrawn in a given quarter (dummy = 1).
  - Alternative: equity-to-assets ratio below 2% (used in robustness checks).

### III. Key empirical findings
- Aggregate liquidity creation evolution (2000–2007):
  - LC1 rises from 22 to 30 percent of assets.
  - LC2 hovers around 18 to 21 percent of assets.
  - LC1 consistently larger than LC2 and more volatile.
- Baseline panel logit results (four quarterly lags before failure):
  - LC_90-100%:
    - Coefficient positive and significant at the 1% level across all quarters prior to failure and using both LC1 and LC2.
    - Interpretation: banks with liquidity creation ratios exceeding the 90th percentile in the system in the quarters prior to failure are more likely to fail.
  - LC_0-10%:
    - Coefficient positive and significant in about half of the baseline estimations — limited evidence that very low liquidity creation also increases failure probability.
  - Control variables:
    - Size: negative and significant in all estimations.
    - ROA: negative and significant in all estimations.
    - Household Income Growth and Small Business Growth: not statistically significant in baseline.
- Robustness checks (summary of alternative specifications and outcomes):
  - Longer lags (15, 18, 21, 24 months):
    - LC_90-100% remains positive and significant across estimations.
    - LC_0-10% positive and significant only at the 21-month lag.
  - Alternative failure definition (equity-to-assets < 2%):
    - LC_90-100% remains positive and significant in all estimations.
    - LC_0-10% not consistently significant.
  - Moscow-only sample:
    - LC_90-100% positive and significant across estimations.
    - Low liquidity creation has no impact.
  - Domestic private banks only:
    - LC_90-100% positive and significant at the 1% level.
    - LC_0-10% positive and significant in most estimations.
  - Deposit insurance (dummy from beginning of 2005):
    - Deposit Insurance coefficient not significant.
    - LC_90-100% remains positive and significant across estimations.
    - Some evidence that LC_0-10% associates with higher failure probability in these specifications.
  - Bank concentration (regional Herfindahl):
    - LC_90-100% positive and highly significant across estimations.
    - LC_0-10% generally not significant.
    - Herfindahl coefficient significant and negative — consistent with concentration reducing probability of bank failure in Russia.
- Synthesis of findings:
  - Consistent and robust evidence supports the High Liquidity Creation Hypothesis (HLCH): banks in the top tail (above the 90th percentile) of liquidity creation face a significantly higher probability of failure under normal economic conditions.
  - Evidence for a positive link between very low liquidity creation and failures is weaker and sensitive to specification.

### IV. Policy implications and recommendations
- Regulatory monitoring:
  - Screen banks by liquidity creation ranking in the system rather than implying a single absolute cutoff.
  - Place high liquidity creators (those above the 90th percentile in a given quarter) on watch lists for enhanced supervision.
- Early warning systems:
  - Incorporate liquidity creation measures (LC1 and LC2 normalized by total assets) into early warning frameworks to identify banks whose core liquidity-creating activity may elevate failure risk.
- Prudential actions:
  - Consider interventions to reduce excessively high liquidity creation to limit the probability of failure and potential taxpayer costs from resolution.
  - For high liquidity creators, regulators may prefer preventive oversight or orderly wind-downs to limit systemic spillovers and preserve aggregate liquidity creation.
- Deposit insurance and safety nets:
  - Introduction of deposit insurance in Russia (beginning of 2005) did not show a statistically significant effect on failure probability in this analysis; policymakers should weigh deposit insurance’s two opposing effects (reducing depositor runs vs. encouraging risk-taking).
- Market structure:
  - Given the negative and significant relation between regional concentration (Herfindahl) and failure probability, authorities should account for local concentration dynamics when assessing systemic vulnerability.

### V. Conclusion (concise)
- The paper documents robust evidence that high liquidity creation by individual banks increases the probability of bank failure in Russia during 2000–2007, a period without major exogenous shocks to the financial system.
- The HLCH complements WFH and LSH by focusing on the core liquidity-creating activity of banks as an endogenous source of fragility.
- Policy relevance: rank-based screening of liquidity creators, integrating liquidity creation into supervision and early warning systems, and targeted oversight of high liquidity creators to prevent bank failures and limit taxpayer losses.

*Source: 10. Liquidity Creation and Bank Failures: Effect of Bank Concentration (chapter content). PDF: _wp15103 - 10. Liquidity Creation and Bank Failures: Effect of Bank Concentration.*

### REFERENCES

### _wp15103 - REFERENCES

### Literature and empirical background
- The references list cites foundational and recent work on bank runs, liquidity creation, bank failures, regulation, and crises. Examples include:
  - Allen, F. and D. Gale, 1998, Optimal Financial Crises, Journal of Finance 53, 1245-84.
  - Diamond, D. and P. Dybvig, 1983, Bank Runs, Deposit Insurance, and Liquidity, Journal of Political Economy 91, 401-19.
  - Berger, A. and C. Bouwman, 2009, Bank Liquidity Creation, Review of Financial Studies 22, 3779-3837.
  - Demirgüç-Kunt, A., and E. Detragiache, 1998, The Determinants of Banking Crises in Developing and Developed Countries, IMF Staff Papers 45 (1), 81-109.
  - Vazquez, F. and P. Federico, 2012, Bank Funding Structures and Risk: Evidence from the Global Financial Crisis, IMF Working Paper 12/29.
- The bibliography covers theoretical models (e.g., Bryant 1980; Diamond & Rajan 2000, 2001, 2002, 2005), historical accounts (e.g., Kindleberger 1978; Calomiris 2007), cross-country empirical studies (e.g., Caprio & Klingebiel 1996), and papers on bank-level failure prediction and hazard models (e.g., Meyer & Pifer 1970; Molina 2002; Whalen 1991).

### Table 1 — Liquidity creation measures (classification and weights)
- LC1: CATEGORY MEASURE
  - Asset-side:
    - Illiquid assets (1/2): Loans to firms; Other assets; Loans to individuals
    - Semi-liquid assets (0): Interbank loans; Loans to government
    - Liquid assets (-1/2): Correspondent accounts with other banks; Government securities (incl. securities issued by regions and municipalities); Investments in promissory notes
  - Liability-side:
    - Liquid liabilities (1/2): Debt securities issued (promissory notes); Debt securities issued (deposit and saving certificates, bonds); Other liabilities
    - Semi-liquid liabilities (0): Claims of non-bank sector: settlement accounts (firms, households, government); Claims of non-bank sector: term deposits accounts (firms, households, government)
    - Illiquid liabilities and capital (-1/2): Capital; Claims of banks
- LC2: MATURITY MEASURE
  - Asset-side (maturity distinction for interbank loans):
    - Illiquid assets (1/2): Interbank loans (maturity more than 1 year); Loans to firms; Other assets; Loans to individuals
    - Semi-liquid assets (0): Interbank loans (maturity more than a week and less than 1 year); Loans to government
    - Liquid assets (-1/2): Interbank loans (maturity less than a week); Correspondent accounts with other banks; Government securities (incl. securities issued by regions and municipalities); Investments in promissory notes
  - Liability-side (maturity distinction):
    - Liquid liabilities (1/2): Liabilities with maturity lower than 90 days
    - Semi-liquid liabilities (0): Liabilities (term deposits and debt securities) with maturity less than 1 year
    - Illiquid liabilities and capital (-1/2): Liabilities (term deposits, debt securities) with maturity more than 1 year and overdue liabilities and liabilities with uncertain term to maturity; Current and corresponding accounts; Capital

### Table 2 — Development of the main variables (selected quarterly observations)
- LC1 and LC2 are expressed as proportion of total assets. Size denotes total assets in millions of rubles. Number of failures counts failed banks with data available four quarters before failure.
- Selected rows (Obs., LC1/assets (mean in %), LC2/assets (mean in %), Size, Number of failures):
  - 2000q1: 1214; 20.91; 17.50; 1 280.7; 14
  - 2000q2: 1222; 21.30; 17.89; 1 389.5; 7
  - 2000q3: 1227; 22.49; 18.84; 1 597.7; 9
  - 2006q1: 845; 30.35; 18.39; 11 127.6; 7
  - 2006q2: 850; 30.11; 18.82; 12 175.8; 14
  - 2007q4: 983; 28.60; 18.63; 19 276.4; 4

### Table 3 — Descriptive statistics: high liquidity creators vs other banks
- Definition: High liquidity creators = banks with liquidity creation measure above the 90th percentile of the distribution.
- Panel A (Category measure, LC1) — selected comparisons (all variables in percent except Size and Z-score). Values shown as "High Liquidity Creators" then "Other Banks":
  - Equity/total assets: 13.45 *** 29.79
  - Size (ln of assets): 6.83 *** 5.76
  - Loans/total assets: 59.49 *** 46.48
  - Deposits/total assets: 52.24 *** 51.50
  - Liquid assets/total assets: 10.94 *** 26.44
  - Liquid liabilities/assets: 50.41 *** 37.49
  - Z-score: 20.05 *** 33.42
  - Loans to firms/assets: 54.59 *** 38.90
  - Loans to households/assets: 4.32 *** 6.95
  - Households deposits/assets: 12.24 *** 15.67
  - Firm deposits/assets: 37.14 *** 34.41
- Panel B (Maturity measure, LC2) — selected comparisons:
  - Equity/total assets: 14.72 *** 29.63
  - Size (ln of assets): 6.08 *** 5.84
  - Loans/total assets: 55.37 *** 46.94
  - Deposits/total assets: 63.06 *** 50.33
  - Liquid assets/total assets: 13.14 *** 26.18
  - Liquid liabilities/assets: 48.78 *** 37.68
  - Z-score: 20.03 *** 33.40
  - Loans to firms/assets: 49.89 *** 39.43
  - Loans to households/assets: 4.83 *** 6.89
  - Households deposits/assets: 16.85 *** 15.17
  - Firm deposits/assets: 42.96 *** 33.78

### Table 4 — Liquidity creation and bank failures: baseline logit (random effects)
- Dependent variable: bank failure (license revoked = 1).
- LC measures entered as percentile dummies. Marginal effects reported. Quarter-specific columns for 1 to 4 quarters before failure.
- Key consistent results (selected coefficients with standard errors in square brackets):
  - LC_90_100%:
    - LC1/Assets, 1 quarter: 1.355*** [0.205]
    - LC1/Assets, 4 quarters: 1.093*** [0.225]
    - LC2/Assets, 1 quarter: 1.259*** [0.212]
    - LC2/Assets, 4 quarters: 1.122*** [0.226]
  - Size:
    - LC1/Assets, 1 quarter: -0.255*** [0.054]
    - LC1/Assets, 4 quarters: -0.294*** [0.083]
    - LC2/Assets, 1 quarter: -0.227*** [0.062]
    - LC2/Assets, 4 quarters: -0.299*** [0.082]
  - ROA:
    - LC1/Assets, 1 quarter: -6.009*** [0.962]
    - LC1/Assets, 4 quarters: -4.535*** [1.484]
    - LC2/Assets, 1 quarter: -6.278*** [1.020]
    - LC2/Assets, 4 quarters: -4.190*** [1.595]
- Sample sizes and fit:
  - Observations range: 35,287 to 34,586 across quarter specifications.
  - Nb. of banks: 1386 or 1385 depending on quarter.
  - LogLikelihood examples: -1064.607 (LC1/1 quarter), -1320.691 (LC1/4 quarters).

### Table 5 — Liquidity creation and bank failures: longer lags (months before failure)
- Columns report results for 15, 18, 21, 24 months before failure for LC1 and LC2.
- Notable coefficients:
  - LC_90_100%:
    - LC1/Assets, 24 months: 0.846*** [0.230]
    - LC2/Assets, 15 months: 0.820*** [0.238]
    - LC2/Assets, 24 months: 0.748*** [0.268]
  - Size:
    - LC1/Assets, 15 months: -0.253*** [0.077]
    - LC1/Assets, 24 months: -0.124** [0.057]
    - LC2/Assets, 24 months: -0.164** [0.074]
  - ROA: coefficients less precisely estimated at long horizons (examples):
    - LC1/Assets, 15 months: -3.411 [2.217]
    - LC2/Assets, 18 months: -3.768** [1.843]
- Observations:
  - 15 months: 33,386; 24 months: 30,207.
  - Nb. of banks declines from 1,374 to 1,331 as lag increases.

### Table 6 — Alternative failure definition (equity/total assets < 2%)
- Dependent variable redefined as equity/total assets below 2%.
- Key results (quarters before failure):
  - LC_90_100%:
    - LC1/Assets, 1 quarter: 1.906*** [0.340]
    - LC1/Assets, 3 quarters: 2.056*** [0.356]
    - LC2/Assets, 1 quarter: 1.498*** [0.343]
    - LC2/Assets, 4 quarters: 1.055*** [0.376]
  - Size:
    - LC1/Assets, 4 quarters: 0.128* [0.075]
    - LC2/Assets, 4 quarters: 0.195*** [0.074]
  - ROA:
    - LC1/Assets, 1 quarter: -2.984*** [0.886]
    - LC1/Assets, 4 quarters: -3.958** [1.629]
    - LC2/Assets, 4 quarters: -4.665*** [1.591]
- Observations: 35,287 down to 34,586; Nb. of banks: 1,386/1,385.

### Table 7 — Moscow banks only
- Estimations limited to banks located in Moscow. Regional variables omitted.
- Key results (quarters before failure):
  - LC_90_100%:
    - LC1/Assets, 1 quarter: 1.229*** [0.248]
    - LC1/Assets, 4 quarters: 1.096*** [0.215]
    - LC2/Assets, 1 quarter: 1.305*** [0.250]
    - LC2/Assets, 4 quarters: 1.083*** [0.220]
  - Size:
    - LC1/Assets, 1 quarter: -0.412*** [0.065]
    - LC1/Assets, 3 quarters: -0.320*** [0.057]
    - LC2/Assets, 4 quarters: -0.259*** [0.053]
  - ROA:
    - LC1/Assets, 1 quarter: -6.112*** [1.238]
    - LC2/Assets, 1 quarter: -6.126*** [1.265]
- Observations: ~16,240 to 15,776; Nb. of banks: 686–688.

### Table 8 — Domestic private banks only
- Subsample: domestic private banks.
- Key results mirror baseline:
  - LC_90_100%:
    - LC1/Assets, 1 quarter: 1.361*** [0.207]
    - LC1/Assets, 4 quarters: 1.090*** [0.217]
    - LC2/Assets, 1 quarter: 1.247*** [0.211]
    - LC2/Assets, 4 quarters: 1.111*** [0.241]
  - Size:
    - LC1/Assets, 1 quarter: -0.223*** [0.055]
    - LC2/Assets, 4 quarters: -0.338*** [0.072]
  - ROA:
    - LC1/Assets, 1 quarter: -6.027*** [0.978]
    - LC2/Assets, 1 quarter: -6.261*** [1.030]
- Observations: 33,097 to 32,450; Nb. of banks ~1,322–1,323.

### Table 9 — Effect of deposit insurance (dummy = 1 after implementation in 2004)
- Deposit insurance included as a dummy variable from 2004 onward.
- Key results:
  - LC_90_100%:
    - LC1/Assets, 1 quarter: 1.354*** [0.205]
    - LC2/Assets, 1 quarter: 1.244*** [0.201]
  - Deposit insurance coefficient examples:
    - LC1/Assets, 1 quarter: -0.220 [0.478]
    - LC1/Assets, 3 quarters: -0.433 [0.488]
    - LC2/Assets, 2 quarters: 0.280 [0.378]
    - One column shows Deposit insurance with standard error [0.000] for 4 quarters (reported as 0.000 in table).
  - Size and ROA remain with expected signs (Size negative; ROA negative and significant).
- Observations: 35,287 down to 34,586; Nb. of banks: 1,386/1,385.

### Table 10 — Effect of regional bank concentration (Herfindahl index)
- Herfindahl index included as regional bank concentration measure.
- Key results:
  - LC_90_100%:
    - LC1/Assets, 1 quarter: 1.370*** [0.197]
    - LC2/Assets, 1 quarter: 1.290*** [0.196]
  - Herfindahl index:
    - LC1/Assets, 1 quarter: -3.933*** [0.659]
    - LC1/Assets, 4 quarters: -3.464*** [0.558]
    - LC2/Assets, 1 quarter: -4.018*** [0.663]
    - LC2/Assets, 4 quarters: -3.592*** [0.562]
  - Size and ROA again show negative marginal effects on failure probability:
    - Size, LC1/1 quarter: -0.360*** [0.051]
    - ROA, LC1/1 quarter: -5.635*** [0.726]
- Observations: 35,287 to 34,586; Nb. of banks: 1,386/1,385.

### Cross-cutting empirical findings (from the tables)
- Banks in the top decile of liquidity creation (LC_90_100%) exhibit robustly positive and statistically significant marginal effects on the probability of failure across specifications, measures (LC1 and LC2), samples (full, Moscow-only, domestic private), and alternative failure definitions. Representative coefficients are in the range:
  - LC_90_100%: approximately 1.229 to 2.056 depending on specification and failure definition, with many estimates significant at the 1% level.
- Size (log assets) is consistently negatively associated with failure probability in most specifications:
  - Representative marginal effects: e.g., -0.255*** [0.054], -0.360*** [0.051], -0.412*** [0.065] in various tables.
- Profitability (ROA) is strongly and negatively related to failure probability across specifications:
  - Representative marginal effects: e.g., -6.009*** [0.962], -6.278*** [1.020], -6.112*** [1.238].
- Regional factors and institutional variables alter magnitudes:
  - Herfindahl index (regional concentration) has large negative marginal effects (e.g., -3.933*** [0.659]), suggesting higher regional concentration reduces failure probability.
  - Deposit insurance coefficients are small and imprecisely estimated in these specifications (examples reported above).

*Content unit: _wp15103 - REFERENCES (extracted from source PDF)*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15103.pdf_