## _wp11258

## Source details

**Canonical URL:** [_wp11258](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11258.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11258.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11258.pdf.json)

---

### I. Introduction — central banks, monetization, and emerging markets
- Recent experience in advanced economies: central banks provided aggregate liquidity, then large financial support to illiquid and insolvent banks, and reserve money eventually increased; despite extensive monetization, these policies did not fuel inflation pressures and exchange rate behavior varied (U.S. dollar appreciated; British pound and euro depreciated gradually).
- Argument for emerging markets: large-scale monetization in emerging markets and developing countries can cause further macroeconomic unrest and increase the chances of a simultaneous currency crisis.
- Scope: analysis of episodes of financial turmoil in 16 Latin American countries during 1995–2007, macroeconomic repercussions, empirical analysis, and lessons.
- Caveats:
  - Financial assistance from multilateral institutions is not factored explicitly.
  - Soundness of the banking system is not incorporated.
  - Many Latin American central banks lacked credibility during most of the period.

### II. Stylized facts on Latin American banking crises (1990s–2000s)
- Frequency and context:
  - Latin America had 28 systemic banking crises from 1970 to 2007.
  - Since 1990, only Chile and Panama were immune to financial instability events; several countries were affected more than once (Argentina, Bolivia, Dominican Republic, Ecuador, Guatemala, Paraguay).
- Typical origin story: financial liberalization → capital inflows → declining real interest rates → currency appreciation → credit booms → lax supervision → external shocks/capital outflows → liquidity and solvency problems.
- Features increasing crisis propensity:
  - Emerging market status, financial dollarization, weak fiscal stance (public debt denominated in foreign currencies).

### III. Central banks’ menu of responses in Latin America
- Initial responses: intensive use of standing facilities and lender-of-last-resort (LLR) short-run loans at penalized interest rates, subject to collateral requirements.
- Broader interventions as crises evolved:
  - (i) Supply enhanced LLR—in amounts that exceeded the impaired banks’ equity;
  - (ii) Grant financial assistance to support bank resolution and restructuring (issuing securities swapped for nonperforming assets; P&A operations);
  - (iii) Fund payments of deposit insurance/guarantees and finance deposit withdrawals.
- Institutional context: weak early corrective action frameworks, poorly funded or absent deposit insurance, lack of effective resolution instruments led to heavy reliance on central bank money.
- Contrasting cases with limited monetization: Argentina (1995), Colombia (1999), Peru (1999), Brazil, Nicaragua—relied more on bank resolution measures and limited LLR; IMF support and strong macroeconomic foundations helped in some cases.

### IV. Macroeconomic repercussions of monetization in Latin America
- Balance-sheet expansions:
  - Some central banks more than doubled their balance sheets in the first two years of their crisis (index evidence in Table 1).
- Peak central bank credit (shares of GDP at the peak of crises):
  - Argentina (2002): monetary assistance exceeded 8 percent of GDP.
  - Dominican Republic (2003): central bank assistance reached nearly 20 percent of GDP.
  - Ecuador (end-1999): BCE injected nearly 12 percent of GDP.
  - Mexico (by end-1994 via Bank of Mexico and FOBAPROA): close to 10 percent of GDP.
  - Uruguay (mid-2002): central bank assistance exceeded 10 percent of GDP.
  - Venezuela (1995): near 10 percent of GDP at some point.
- Short-run monetary policy constraint:
  - Mopping up excess liquidity would have required very high interest rates; interbank market segmented; reserve money boomed except where FX loans/interventions reduced reserves.
  - Short-term interest rates increased despite reserve money expansion.
- Exchange rate and reserves dynamics:
  - Large depreciations occurred in systemic crises (examples: Ecuador 1999 local currencies depreciated more than 50 percent in three months; Argentina 2002 U.S. dollar value quadrupled relative to peso).
  - Central bank FX interventions reduced international reserves; pegs were abandoned in several cases (Argentina 2002, Uruguay 2002, Ecuador 1999, Mexico 1994).
  - Vicious cycle: monetization → depreciation expectations → higher interest rates → worsened bank solvency → more central bank support → further depreciation.
  - Administrative measures used in some cases (capital controls in Venezuela; limits on deposits in Argentina, Uruguay, Ecuador).
  - Multilateral support helped in some crises (Dominican Republic 2003, Mexico in mid-1990s).
- Inflation and growth:
  - Monetization often undermined price stability via exchange-rate pass-through.
  - Short-run growth effects typically negative: majority of large-crisis countries experienced real GDP contractions of at least 3 percent.
  - Channels of growth impact:
    - Exchange rate channel: depreciation can improve competitiveness and exports in medium term.
    - Interest rate / credit channel: higher interest rates and credit hoarding reduce credit availability and aggregate demand; in dollarized economies depreciation weakens unhedged borrowers’ net wealth and banks’ balance sheets.

### V. Empirical analysis (panel regressions, 1995–2007)
- Sample: 16 Latin American countries; annual observations 1995–2007; num. obs. = 208.
- Key variables:
  - Monetization measure: change in central bank claims on depositary institutions as a ratio of GDP (Δ Claims).
  - External variables: % change in international reserves; % change in nominal exchange rate (local currency per U.S. dollar; positive = depreciation).
- Summary descriptive statistics (annual observations, 1995–2007; num. obs. = 208):
  - Full sample:
    - Mean Δ Credit to banks / GDP: -0.2
    - % Δ in international reserves: 17.4
    - % Δ in nominal exchange rate: 10.2
    - Median Δ Credit to banks / GDP: -0.1
    - Median % Δ in international reserves: 12.9
    - Median % Δ in nominal exchange rate: 5.0
    - Standard deviations: 1.7 (Δ Credit/GDP), 35.2 (% Δ reserves), 23.6 (% Δ exchange rate)
    - Minimums: -11.2 (Δ Credit/GDP), -75.2 (% Δ reserves), -27.8 (% Δ exchange rate)
    - Maximums: 6.7 (Δ Credit/GDP), 215.3 (% Δ reserves), 206.5 (% Δ exchange rate)
  - Episodes of credit expansion (Δ Claims > 0; num. obs. = 66):
    - Mean Δ Credit to banks / GDP: 1.0
    - Mean % Δ in international reserves: 13.4
    - Mean % Δ in nominal exchange rate: 15.2
    - Medians: 0.4 (Δ Credit/GDP), 6.7 (% Δ reserves), 5.6 (% Δ exchange rate)
    - Standard deviations: 1.5 (Δ Credit/GDP), 45.9 (% Δ reserves), 32.4 (% Δ exchange rate)
    - Minimums: 0.0 (Δ Credit/GDP), -75.2 (% Δ reserves), -8.7 (% Δ exchange rate)
    - Maximums: 6.7 (Δ Credit/GDP), 215.3 (% Δ reserves), 206.5 (% Δ exchange rate)
- Panel unit-root tests (Levin-Lin-Chu):
  - All series are I(0):
    - ΔNXR: -8.7*** → I(0)
    - Δ Claims: -9.5*** → I(0)
    - IR/M2: -3.0*** → I(0)
    - Debt/GDP: -5.3*** → I(0)
    - TT: -1.9** → I(0)
    - Significance: * p<0.1, ** p<0.05, *** p<0.01
- Econometric model:
  - ΔNXRit = α + β ΔCit + γ Zit + μi + λt + νit
    - ΔNXR = percentage change in nominal exchange rate
    - ΔC = change in central bank credit to the banking system as percent of GDP (Δ Claims)
    - Z = controls: lagged public debt/GDP, lagged IR/M2, lagged dependent variable, terms of trade (TT)
- Main regression results (fixed effects, full sample; Table 5):
  - Coefficients for Δ Claims are positive and statistically significant (at least at the 5 percent level) across FE specifications and with controls.
  - Numerical interpretation: on average, an annual increase of 1 percentage point in the ratio of credit to GDP provided by central banks to banks leads domestic currencies to depreciate about 2.5 percent in one year.
  - Terms of trade are significant determinants of exchange rate fluctuations; fiscal conditions (debt/GDP) have expected effects in some specifications.
  - R-squared values around 0.26–0.31 in FE specifications (num. obs. 207–208; num. countries 16).

### VI. Robustness checks and alternative specifications
- System GMM:
  - Addresses endogeneity and dynamic panel bias; internal instruments used; terms of trade treated as strictly exogenous.
  - Diagnostics: ar1: 0.0367; ar2: 0.468; sargan: 0.826.
  - System GMM ∆ Claims coefficient: 2.64** (Table 6, column 1).
- CLSDV and alternative estimators:
  - CLSDV ∆ Claims = 2.62***.
  - Bias-corrected LSDV estimator discussed (Bruno, 2005) with caveat that it does not address endogeneity.
- Alternative dependent variable:
  - NEER used instead of bilateral exchange rate; main conclusion maintained; central bank credit coefficients remain statistically significant (Table 6, column 3).
- Alternative monetization definition:
  - Percent increase in central banks’ claim on banks relative to the average of the previous three years; results similar (not reported).
- Sub-sample: periods of credit expansion
  - Panel regression restricting sample to credit expansion (Table 7):
    - FE/LSDV/POLS specifications show ∆ Claims coefficients around 8.40*** to 9.03*** depending on estimator.
    - Interpretation: pressures on the exchange rate rise as central bank financial assistance to impaired banks escalate.
    - Magnitude example: a 5 percentage point increase in credit to banking system would induce on average a 45 percent decline in exchange rate.

### VII. Estimation methods and diagnostics (additional)
- Diagnostic summary:
  - "Robust standard errors in parentheses"
  - Significance notation: "*** p<0.01, ** p<0.05, * p<0.1"
  - Panel coverage: "No. of countries 16 16 16"
- Reported R-squared snippets:
  - "0.3                        R-squared                        0.52                        0.42                        0.66"

### VIII. Key empirical findings — concise list
- Monetization (Δ Claims) is positively and significantly associated with nominal exchange rate depreciation across specifications.
- Quantitative effect (full sample): 1 percentage point increase in Δ Claims → about 2.5 percent depreciation in one year.
- System GMM and CLSDV confirm similar magnitudes: 2.64** and 2.62*** respectively.
- Effects are substantially larger during credit-expansion/monetization episodes: ∆ Claims coefficients ≈ 8.40*** to 9.03***.
- A 5 percentage point increase in credit to the banking system is associated with an average 45 percent decline in the exchange rate in the credit-expansion subsample.
- Unit-root tests indicate series are stationary (I(0)).

### IX. Policy implications and lessons distilled
- Main policy message:
  - Central banks in emerging markets and developing countries should be cautious when using central bank money to cope with financial crises—large-scale monetization tended to fuel macroeconomic and financial instability and raised the probability of a simultaneous currency crisis in Latin America during 1995–2007.
- Institutional prerequisites to limit monetization costs:
  - Strengthen legal and institutional frameworks for early corrective actions and bank resolution (P&A, restructuring).
  - Ensure credible and funded deposit insurance mechanisms.
  - Limit the amount and maturity of LLR where appropriate; central bank laws that constrain LLR can help avoid excessive monetization.
  - Use central bank interventions as part of a broader financial safety net emphasizing prompt bank resolution rather than prolonged monetization.
- Complementary actions when central bank support is unavoidable:
  - Coordinated use of FX reserves and fiscal/structural measures; seek timely multilateral financial support.
  - Preserve central bank credibility to reduce uncertainty and capital outflow pressures.
  - Avoid excessively relaxing collateral quality standards; monitor central bank capital and ensure transparent replenishment when losses materialize.

### X. Empirical inventory: selected banking crises in Latin America (1994–2008)
- Selected small crises:
  - Argentina (1995–1996): about 40 small and medium size banks failed or were acquired/merged (~12 percent of the system).
  - Bolivia (1994): Two banks (market share 11 percent of assets) were closed.
  - Bolivia (1999): A small bank (market share 4.5 percent of deposits) was intervened and resolved.
  - Guatemala (1999): Five banks (9 percent market share) received large financial assistance from the central bank (more than 100 percent and up to 467 percent of equity).
  - Paraguay (1995): Four banks (market share ~14 percent of total assets) were intervened and closed.
- Selected large crises:
  - Argentina (2002): 12 private and public banks (40 percent of deposits) received liquidity assistance; bank resolution applied to three foreign banks.
  - Brazil (1994–1995): 18 financial institutions (nearly 35 percent market share) were intervened, liquidated, or placed under special administration; comprehensive restructuring adopted.
  - Ecuador (1998–1999): closures/interventions representing ~20 percent deposits (1998) and close to 40 percent deposits (1999).
  - Mexico (1995): Twelve small and medium-size banks intervened during 1995–1997 and a group of banks covering 80 percent of the system received government support.
  - Uruguay (2002): actions covering about 60 percent market share (deposits).
  - Venezuela (1994–1995): Twelve banks closed or taken over in 1994 and another 4 in 1995, totaling 62 percent market share (deposits).

*Italic source: IMF working paper chapter “6. Monetization of Banking Crises and Economic Growth” (pages and figures as provided in the supplied content).*

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

### _wp11258 - References ...............................................................................................................

### Tables

- 1. Expansion of Central Banks’ Balance Sheets during Banking Crises .................................. 9
- 2. Summary Descriptive Statistics .......................................................................................... 18
- 3. Pair-wise Correlations Between Selected Variables ........................................................... 19
- 4. Panel Unit Root Tests—Levin-Lin-Chu ............................................................................. 21
- 5. Panel Regressions of Changes in Nominal Exchange Rates Due to Changes in  
  Central Bank Claims ............................................................................................................ 22
- 6. Additional Robustness Checks ............................................................................................ 24
- 7. Panel Regression Restricting the Sample to Periods of Credit Expansion ......................... 24

### Figures

- 1. Monetization of Banking Crises and Base Money .............................................................. 12
- 2. Monetization of Banking crises and Short-Term Interest Rates ......................................... 12
- 3. Monetization of Large Banking Crises and Exchange Rate Depreciation .......................... 14
- 4. Monetizing Banking Crises and International Reserves ..................................................... 14
- 5. Exchange Rate Depreciation and Inflation ......................................................................... 16

*Source: _wp11258 - References (PDF).*

### 6. Monetization of Banking Crises and Economic Growth .................................................... 16

### 6. Monetization of Banking Crises and Economic Growth

### I. Introduction — central banks, monetization, and emerging markets
- Recent experience in advanced economies: central banks provided aggregate liquidity, then large financial support to illiquid and insolvent banks, and reserve money eventually increased; despite extensive monetization, these policies did not fuel inflation pressures and exchange rate behavior varied (U.S. dollar appreciated; British pound and euro depreciated gradually).
- Emerging markets and developing countries may be tempted to emulate advanced-economy policies, but the paper argues large-scale monetization in emerging markets and developing countries can cause further macroeconomic unrest and increase the chances of a simultaneous currency crisis.
- Scope: analysis of episodes of financial turmoil in 16 Latin American countries during 1995–2007, macroeconomic repercussions, empirical analysis, and lessons.
- Caveats highlighted in the paper:
  - Financial assistance from multilateral institutions (most typically from the IMF) to some countries at the time of the crises is not factored explicitly into the paper.
  - Soundness of the banking system is not incorporated in the analysis.
  - During most of the period, many Latin American central banks lacked credibility; credible central banks can better mitigate uncertainty that fosters capital outflows.

### II. Stylized facts on Latin American banking crises (1990s–2000s)
- Frequency and context:
  - Latin America had 28 systemic banking crises from 1970 to 2007.
  - Since 1990, only Chile and Panama were immune to events of financial instability; several countries were affected more than once (Argentina, Bolivia, Dominican Republic, Ecuador, Guatemala, Paraguay).
- Typical origin story: a “boom and bust cycle” driven by financial liberalization, capital inflows, declining real interest rates, currency appreciation, credit booms, and lax supervision; subsequent external shocks and capital outflows produced liquidity and solvency problems.
- Features that made countries more prone to systemic crises:
  - Emerging market status (vulnerability to external shocks),
  - Financial dollarization (facilitating balance sheet mismatches),
  - Weak fiscal stance (public debt denominated in foreign currencies amplified depreciation impacts).

### III. Central banks’ menu of responses in Latin America
- Initial responses: recycling liquidity via intensive use of standing facilities and lender-of-last-resort (LLR) short-run loans at penalized interest rates, subject to collateral requirements.
- Broader interventions as crises evolved (three broad courses):
  - (i) Supply enhanced LLR—in amounts that exceeded the impaired banks’ equity;
  - (ii) Grant financial assistance to support bank resolution and restructuring (e.g., issuing securities swapped for nonperforming assets; P&A operations);
  - (iii) Fund payments of deposit insurance/guarantees and finance deposit withdrawals.
- Central banks often stretched LLR to assist institutions facing solvency problems; collateral quality was sometimes relaxed as crises worsened.
- Lack of institutional frameworks (weak early corrective action frameworks, poorly funded or absent deposit insurance, lack of effective resolution instruments) underpinned heavy reliance on central bank money.
- Contrasting cases where institutional frameworks limited monetization:
  - Argentina (1995), Colombia (1999), Peru (1999), Brazil, Nicaragua—these relied more on bank resolution measures (P&A, restructuring) and limited LLR; IMF support and strong macroeconomic foundations helped in some cases.

### IV. Macroeconomic repercussions of monetization in Latin America
- Balance-sheet expansions:
  - Some central banks more than doubled their balance sheets in the first two years of their crisis (index evidence in Table 1).
  - Selected index numbers (T = pre-crisis year; T+1 and T+2 first and second years of crises) summarized in Table 1 for advanced and Latin American countries (source: IFS and central bank reports).
- Peak central bank credit (examples, stated as shares of GDP at the peak of crises):
  - Argentina (2002): monetary assistance exceeded 8 percent of GDP (mostly to public banks).
  - Dominican Republic (2003): central bank assistance reached nearly 20 percent of GDP.
  - Ecuador (end-1999): BCE injected nearly 12 percent of GDP.
  - Mexico (by end-1994 via Bank of Mexico and FOBAPROA): close to 10 percent of GDP.
  - Uruguay (mid-2002): central bank assistance exceeded 10 percent of GDP.
  - Venezuela (1995): near 10 percent of GDP at some point.
- Short-run monetary policy constraint:
  - Large financial support constrained central banks’ ability to conduct monetary policy because mopping up excess liquidity would have required very high interest rates.
  - Interbank market segmentation: flight to quality and capital outflows; reserve money boomed (Figure 1) except where central banks provided foreign-currency loans or intervened to reduce reserves.
  - Short-term interest rates increased despite reserve money expansion (Figure 2); this raised funding costs and undermined solvency.
- Exchange rate and reserves dynamics:
  - Increased demand for foreign currency led to exchange rate pressures and large depreciations in systemic crises (e.g., local currencies depreciated more than 50 percent in three months in Ecuador 1999; Argentina 2002 saw U.S. dollar value quadruple relative to peso).
  - Central bank FX interventions reduced international reserves (Figure 4); pegs were abandoned in several cases (Argentina 2002, Uruguay 2002, Ecuador 1999, Mexico 1994).
  - Vicious cycle: monetization → depreciation expectations → higher interest rates → worsened bank solvency → more central bank support → further depreciation.
  - Administrative measures sometimes used to break the loop (capital controls in Venezuela; limits on deposits in Argentina, Uruguay, Ecuador).
  - Multilateral support helped in some crises (Dominican Republic 2003, Mexico in mid-1990s).
- Inflation and growth:
  - Monetization often undermined price stability as exchange-rate pass-through pushed up inflation (Figure 5).
  - Short-run growth effects typically negative: the majority of countries suffering large crises experienced real GDP contractions of at least 3 percent (Figure 6).
  - Two channels for growth impact from central bank money injections:
    - Exchange rate channel: depreciation can improve competitiveness and exports in medium term.
    - Interest rate / credit channel: higher interest rates and hoarding of liquidity by sound banks reduce credit availability, damaging aggregate demand; in dollarized economies depreciation hits unhedged borrowers’ net wealth, weakening banks’ balance sheets and aggregate demand.
- Summary observation from the narrative and figures: excessive reliance on central bank money in Latin America was associated with exchange rate depreciation, reserves losses, inflation acceleration, and short-run contractions in real GDP.

### V. Empirical analysis (panel regressions, 1995–2007)
- Sample: 16 Latin American countries (Argentina, Bolivia, Brazil, Colombia, Costa Rica, Dominican Republic, Ecuador, El Salvador, Guatemala, Honduras, Mexico, Nicaragua, Paraguay, Peru, Uruguay, Venezuela); annual observations 1995–2007.
- Key variables and measurement:
  - Monetization measure: change in central bank claims on depositary institutions as a ratio of GDP (Δ Claims, central bank claims on banks/GDP).
  - External variables: % change in international reserves; % change in nominal exchange rate (local currency per U.S. dollar; positive sign = depreciation).
- Summary descriptive statistics (annual observations, 1995–2007; num. obs. = 208):
  - Full sample:
    - Mean Δ Credit to banks / GDP: -0.2
    - % Δ in international reserves: 17.4
    - % Δ in nominal exchange rate: 10.2
    - Median Δ Credit to banks / GDP: -0.1; Median % Δ in international reserves: 12.9; Median % Δ in nominal exchange rate: 5.0
    - Standard deviations: 1.7 (Δ Credit/GDP), 35.2 (% Δ reserves), 23.6 (% Δ exchange rate)
    - Minimums: -11.2 (Δ Credit/GDP), -75.2 (% Δ reserves), -27.8 (% Δ exchange rate)
    - Maximums: 6.7 (Δ Credit/GDP), 215.3 (% Δ reserves), 206.5 (% Δ exchange rate)
  - Episodes of credit expansion (defined where central bank claims on banks increased; num. obs. = 66):
    - Mean Δ Credit to banks / GDP: 1.0
    - Mean % Δ in international reserves: 13.4
    - Mean % Δ in nominal exchange rate: 15.2
    - Medians: 0.4 (Δ Credit/GDP), 6.7 (% Δ reserves), 5.6 (% Δ exchange rate)
    - Standard deviations: 1.5 (Δ Credit/GDP), 45.9 (% Δ reserves), 32.4 (% Δ exchange rate)
    - Minimums: 0.0 (Δ Credit/GDP), -75.2 (% Δ reserves), -8.7 (% Δ exchange rate)
    - Maximums: 6.7 (Δ Credit/GDP), 215.3 (% Δ reserves), 206.5 (% Δ exchange rate)
- Pair-wise correlations (annual observations, 1995–2007; p-values reported in Table 3):
  - Δ in central bank claims on banks / GDP correlated with nominal depreciation (positive) and with real effective depreciation (positive); negative association with growth; correlation patterns stronger during monetization episodes.
- Econometric model (equation (1), country fixed effects and year dummies):
  - ΔNXRit = α + β ΔCit + γ Zit + μi + λt + νit
    - ΔNXR = percentage change in nominal exchange rate
    - ΔC = change in central bank credit to the banking system as percent of GDP (Δ Claims)
    - Z = control variables: lagged public debt/GDP, lagged IR/M2, lagged dependent variable, terms of trade (TT)
- Panel unit-root tests (Levin-Lin-Chu):
  - All series are I(0) (stationary):
    - ΔNXR: -8.7*** → I(0)
    - Δ Claims: -9.5*** → I(0)
    - IR/M2: -3.0*** → I(0)
    - Debt/GDP: -5.3*** → I(0)
    - TT: -1.9** → I(0)
    - (* significant at 10%; ** at 5%; *** at 1%)
- Main regression results (fixed effects and full 1995–2007 sample; Table 5):
  - Coefficients for Δ Claims are positive, statistically significant (at least at the 5 percent level) across FE specifications and remain so when including controls.
  - Numerical interpretation: on average, an annual increase of 1 percentage point in the ratio of credit to GDP provided by central banks to banks leads domestic currencies to depreciate about 2.5 percent in one year.
  - Terms of trade are significant determinants of exchange rate fluctuations (consistent with “commodity currency” hypothesis); fiscal conditions (debt/GDP) have expected effects in some specifications.
  - R-squared values around 0.26–0.31 in FE specifications (number of observations 207–208; number of countries 16).
- Robustness checks:
  - System GMM: addresses endogeneity and dynamic panel bias; internal instruments used (lagged levels and differences), terms of trade treated as strictly exogenous; Sargan test did not reject instrument validity; AR(1) rejected, AR(2) not rejected (residuals serially uncorrelated in levels). System GMM results are broadly similar: ∆ Claims coefficient in System GMM reported as 2.64** (Table 6, column 1).
  - CLSDV and alternative estimators: CLSDV ∆ Claims = 2.62***; other specifications report similar positive and significant coefficients (Table 6).
  - Alternative dependent variables and samples: restricting the sample to periods of credit expansion yields larger coefficients. Panel regression restricting sample to credit expansion (Table 7):
    - FE/LSDV/POLS specifications show ∆ Claims coefficients around 8.40*** to 9.03*** depending on estimator—i.e., much larger effects when focusing only on expansion episodes.
  - Using NEER instead of bilateral exchange rate, different lags and instrument sets produced qualitatively similar results.
- Empirical conclusion:
  - Enhanced monetization of banking crises (increased central bank claims on banks) is statistically associated with exchange-rate depreciation; effects are economically meaningful and stronger during episodes of credit expansion/monetization.

### VI. Policy implications and lessons distilled
- Central banks in emerging markets and developing countries should be cautious when using central bank money to cope with financial crises—large-scale monetization tended to fuel macroeconomic and financial instability and raised the probability of a simultaneous currency crisis in Latin America during 1995–2007.
- Institutional prerequisites to limit monetization costs:
  - Strengthen legal and institutional frameworks for early corrective actions and bank resolution (e.g., P&A, restructuring), and ensure credible and funded deposit insurance mechanisms.
  - Limit the amount and maturity of LLR where appropriate (central bank laws that constrain LLR can help avoid excessive monetization).
  - Use central bank interventions as part of a broader financial safety net that emphasizes prompt bank resolution rather than prolonged monetization.
- Complementary actions to mitigate adverse effects when central bank support is unavoidable:
  - Coordinated use of FX reserves and fiscal/structural measures; seek timely multilateral financial support to strengthen external position and discourage speculative attacks.
  - Preserve central bank credibility to reduce uncertainty and capital outflow pressures.
  - Avoid relaxing collateral quality standards excessively; monitor central bank capital positions and ensure transparent replenishment when losses materialize to preserve operational autonomy.

*Italic source: IMF working paper chapter “6. Monetization of Banking Crises and Economic Growth” (pages and figures as provided in the supplied content).*

### 0.3                        R-squared                        0.52                        0.42                        0.66

### _wp11258 - 0.3                        R-squared                        0.52                        0.42                        0.66

### Estimation methods and diagnostics
- Bias-corrected least-squares dummy variable (LSDV) estimators:
  - "The bias-corrected least-squares dummy variable (LSDV) estimator could outperform the GMM estimators in terms of bias and root mean squared error (Bruno, 2005), especially for panel data with a small number of cross-sectional units."
  - Limitation: "the bias-corrected LSDV estimator does not deal with the potential endogeneity of independent variables."
- Generalized Method of Moments (GMM):
  - Results "are broadly similar to those obtained with system GMM estimator (Table 6 column 2)."
  - Diagnostics reported in the source:
    - ar1: 0.0367
    - ar2: 0.468
    - sargan: 0.826
    - "Robust standard errors in parentheses"
    - Significance notation: "*** p<0.01, ** p<0.05, * p<0.1"
- Panel coverage:
  - "No. of countries 16 16 16"
  - "Source: Staff calculations."
- Estimation choices for irregular panels:
  - For a sub-sample: "the equations are estimated by Pooling OLS and LSDV. Because we have irregular panel data, the usual fixed effects estimator and system GMM estimator may not be appropriate."

### Robustness checks and alternative specifications
- Dependent variable alternative:
  - Use of NEER instead of bilateral exchange rate vis-à-vis U.S. dollar:
    - Rationale: "Some counties in the sample have fixed exchange rate arrangements, for those countries the dependent variable is a constant and it may not fully capture pressures on exchange rate. For those countries, the NEER provides greater variability of the dependent variable."
    - Result: "The main conclusion of the previous analysis is maintained for NEER exchange rate depreciations. The estimated coefficients for central bank credit remain statistically significant (Table 6, column 3)."
- Alternative definition of variable of interest:
  - Monetization of banking crises measured "by the percentage increase in central banks’ claim on banks with respect to the average of the previous three years in order to capture the increase in credit with respect to the trend."
  - "We estimated the full set of regressions for this variable and, once again, the results were similar (not reported)."
  - Note: "The results are available upon request."

### Key empirical findings
- Sample restriction to periods of credit expansion:
  - "For this sub-sample, the coefficient for credit to the banking system increased significantly, to about 9 percent (columns 1 to 3, Table 7)."
  - Interpretation: "These results suggest that pressures on the exchange rate rise as central bank financial assistance to impaired banks escalate."
  - Magnitude example: "According to these estimations, a 5 percentage point increase in credit to banking system would induce on average a 45 percent decline in exchange rate."
- Relationship maintained across specifications:
  - "The estimated coefficients for central bank credit remain statistically significant" across NEER specification and alternative variable definition.

### Policy conclusions and recommendations
- General assessment:
  - "The role that central banks played in the industrial world in the recent financial crisis will probably set a new standard for policies of LLR. Injecting large amounts of money to support troubled financial institutions averted a financial meltdown and later, in tandem with fiscal stimulus, mitigated recession trends, without fueling inflation pressures."
  - Cautionary note for emerging and developing countries: "As opposed to the United States and most European countries, these countries do not issue a reserve currency. This makes them more vulnerable to currency depreciations and volatility in the event of large scale injections of central bank money to cope with banks’ distress."
- Recommended roadmap for emerging and developing countries:
  - "Put bank restructuring and resolution at the forefront."
  - "Design a comprehensive strategy to prevent and manage banking crises in which central banks are part of the policy response."
  - "Impose corrective actions before liquidity and capital shortages become severe, and on implementing bank resolution measures before the crises unfold."
  - "Strengthen financial regulation and supervision to keep the pace of the permanent innovation of financial instruments."
  - Maintain strong macroeconomic fundamentals:
    - "maintaining flexible exchange rates"
    - "keeping public finances and debt in check"
    - "strengthening international reserves"
    - "developing money and capital markets"
    - "reducing financial dollarization"
  - Rationale: "maintaining strong macroeconomic fundamentals is critical to help reducing the uncertainty that would otherwise invade markets’ perceptions and feed financial instability, including attacks on the domestic currency."

### Empirical inventory: Banking Crises in Latin America (1994 – 2008)
- Small crises (selected entries):
  - Argentina (1995–1996): "about 40 small and medium size banks failed or were acquired or merged (almost one third of total banks) representing about 12 percent of the system."
  - Bolivia (1994): "Two banks (with a market share of 11 percent of assets) were closed."
  - Bolivia (1999): "A small bank, with a market share of 4.5 percent of deposits, was intervened and resolved."
  - Guatemala (1999): "Five banks (9 percent market share) received large financial assistance from the central bank (more than 100 percent and up to 467 percent of equity)."
  - Paraguay (1995): "Four banks, with a market share of about 14 percent of total assets, were intervened and closed."
- Large crises (selected entries):
  - Argentina (2002): "12 private and public banks (40 percent of deposits) received liquidity assistance, and bank resolution was applied to three foreign banks that exited the market."
  - Brazil (1994–1995): "18 financial institutions (nearly 35 percent market share) were intervened, liquidated, or placed under the Regime of Special Temporary Administration. A comprehensive program of bank restructuring was adopted."
  - Ecuador (1998): "closure seven banks and the takeover of a large bank, representing together more than 20 percent deposits’ market share."
  - Ecuador (1999): "a number of financial institutions intervened or resolved, with a market share of close to 40 percent of total deposits."
  - Mexico (1995): "Twelve small and medium-size banks were intervened during 1995–1997 and a group of banks hovering 80 percent of the system received government support."
  - Uruguay (2002): "four private banks were temporarily intervened, closed, and merged in a single bank, two large public banks received financial assistance from the central bank, and one large foreign branch was closed, all of which accounted for about 60 percent market share."
  - Venezuela (1994–1995): "Twelve banks were closed or taken over by the State in 1994 and another 4 in 1995, totaling 62 percent market share (measured by deposits)."

*Source: Staff calculations; selected excerpts from the provided IMF working paper content.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11258.pdf_