## wp17170

## Source details

**Canonical URL:** [wp17170](https://www.imf.org/-/media/files/publications/wp/2017/wp17170.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp17170.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp17170.pdf.json)

---

### Major findings and conclusions
- Macroprudential policy measures targeted at the composition of banks’ funding (liquidity instruments) or the indebtedness of specific sectors (sectoral instruments) often have statistically significant cross-border bank credit spillover effects; measures targeted at the level or quality of banks’ capital buffers (capital instruments) generally do not.
- Empirical evidence of cross-border macroprudential policy spillovers:
  - Stronger for tightening than for loosening measures.
  - Regionally concentrated (larger estimated spillovers in the European subsample).
- Event study and panel regression analyses:
  - Indicate statistically significant credit leakage and reallocation effects for liquidity and sectoral tightening measures.
  - Find little or no evidence of spillovers from capital instrument tightening, possibly because some capital measures are imposed on a consolidated basis.
- Structural model (GFM) simulations:
  - Indicate sectoral policy shocks are generally small worldwide but regionally concentrated and occasionally economically significant.
  - Indicate countercyclical capital buffer adjustments have the potential to generate sizeable regional spillovers.
- Estimated model fit:
  - The baseline panel regression explains a small proportion of variation in directional bilateral cross-border bank credit growth, with R-squared values of about 0.025 for the full sample and 0.050 for the European country subsample.

### Data, coverage, and indices
- Panel coverage:
  - 64 countries over the period 2000Q1 to 2015Q1.
  - Panel dimensions reported: 64N countries and 61T time periods, implying a total of 3,904NT observations.
- Bank credit measures:
  - Domestic bank credit, cross-border bank lending and borrowing, and directional bilateral cross-border bank credit from the BIS International Banking Statistics.
  - Locational cross-border bank credit data are used (based on residency principle).
  - Bank assets in the form of loans are the focus; securities and other claims excluded.
- Macroprudential policy database:
  - New database covering 11 frequently used instruments over 2000Q1 to 2015Q1.
  - Functional classification of the 11 instruments:
    - Capital instruments: time-varying capital requirements, dynamic loan-loss provisions, profit distribution restrictions, countercyclical capital buffers (countercyclical capital buffer adjustments analyzed by structural methods only).
    - Liquidity instruments: loan-to-deposit ratio limits, maturity mismatch limits, net open foreign exchange position limits, reserve requirements.
    - Sectoral instruments: loan-to-value ratio limits, debt-service-to-income ratio limits, debt-to-income or loan-to-income ratio limits, quantitative loan restrictions.
  - Indices:
    - Country-specific, time-varying indicator indices that equal one if at least one instrument in the given category is tightened or loosened at least once during the quarter, and zero otherwise.
    - Indices capture direction but not size of instrument adjustments.
- Special samples and periods:
  - Global Financial Crisis (GFC) period defined as 2008Q1 to 2012Q4.
  - Excluding the GFC period (2008Q1–2012Q4) yields broadly consistent event study results; exclusion almost halves the number of tightening/loosening events under consideration.
  - CESEE subsample: Bulgaria, Croatia, the Czech Republic, Hungary, Poland, Romania, and the Slovak Republic.

### Event study methodology and key empirical signals
- Identification:
  - Nonparametric event study uses only timing and sign restrictions.
  - Persistent spillover events require sign restrictions to be satisfied for at least two consecutive quarters on a year-over-year basis.
- Definitions:
  - Credit leakage: domestic tightening (loosening) associated with increased (decreased) borrowing from foreign banks.
  - Credit reallocation: domestic tightening (loosening) associated with increased (decreased) cross-border lending by domestic banks to foreign borrowers.
- Event study results (summary):
  - Tightening measures using liquidity or sectoral instruments are associated with:
    - Lower domestic bank credit growth.
    - Higher growth in cross-border borrowing from foreign banks.
    - Higher growth in domestic banks’ cross-border lending to foreign borrowers.
  - Little or no evidence of spillovers from tightening measures using capital instruments.
  - Little or no evidence of cross-border spillovers from loosening measures overall (with some exceptions when excluding GFC).
- Magnitude signals from event study medians:
  - On average, cross-border bank borrowing or lending grew 2 to 4 percentage points faster within two quarters of macroprudential policy tightening using liquidity or sectoral instruments (median changes in quarterly year-over-year growth rates).
  - Cross-border bank borrowing or lending grew 2 to 5 percentage points slower within two quarters after macroprudential policy tightening using capital instruments (median changes).

### Panel regression model specification, estimation, and inference
- Theoretical framing:
  - Bilateral cross-border bank credit modeled via a gravity-type relationship: proportional to recipient demand and source supply, inversely proportional to financial distance.
- Reduced-form estimation equation includes:
  - Differences in recipient and source domestic credit growth, nominal output growth, nominal bank lending rate changes, real house price index growth.
  - Differences in financial market development indices, capital flow restriction intensity indices, exchange rate flexibility indices.
  - Cumulated financial crisis indicator and cumulated instrument-specific macroprudential tightening/loosening indices.
  - Pairwise fixed effects and time fixed effects; likelihood ratio tests support their inclusion.
- Estimation details:
  - Annual frequency over 2000–2014 (dataset converted to annual).
  - Dependent variable winsorized at the 5 percent level to remove outliers.
  - Instrumental variables used where required:
    - Domestic bank credit growth in recipient and source countries treated as endogenous.
    - Instruments: growth rate of nominal output, change in nominal bank lending interest rate, growth rate of real house price index (in recipient and source countries).
    - Validity checks performed: Davidson and MacKinnon exogeneity test, Kleibergen and Paap underidentification test, Hansen overidentification test.
  - Dynamic specification rejected due to overidentification test failures for reasonable lag orders; static specification favored.
- Panel regression inference (hypothesis test summaries):
  - Strong evidence (1 percent level) of credit leakage effects from tightening in recipient countries using liquidity instruments (higher growth in cross-border borrowing from foreign banks).
  - Moderate evidence (5 percent level) of credit reallocation effects from tightening of sectoral instruments in source countries (higher growth in cross-border lending by domestic banks to foreign borrowers).
  - Moderate evidence that loosening measures in source countries using liquidity and sectoral instruments are associated with lower growth in domestic bank lending to foreign borrowers.
  - Little or no evidence of leakage or reallocation effects from tightening or loosening measures using capital instruments.
- European subsample results:
  - Broadly consistent with full sample but quantitatively larger: estimated credit leakage and reallocation effects from liquidity and sectoral tightening measures are about 50 and 100 percent larger, respectively, relative to full sample estimates.
- Interactive model specification:
  - Interactions of macroprudential tightening/loosening indices with capital flow restriction intensity indices show:
    - Strong evidence that leakage effects from liquidity tightening are larger in recipient countries with more open capital accounts.
    - Weak evidence that reallocation effects from sectoral tightening are larger in source countries with more open capital accounts.
    - Strong evidence that reallocation effects from sectoral loosening are larger in source countries with more open capital accounts.
    - Little or no evidence that capital instrument spillovers are amplified by more open capital accounts.
- Model fit caveat:
  - Baseline model R-squared about 0.025 (full sample) and 0.050 (European subsample), implying most variation remains unexplained.

### Robustness checks and caveats
- Robustness exercises:
  - Excluding the GFC period (2008Q1–2012Q4).
  - Focusing on CESEE subsample.
  - Interactive specification with capital account openness.
- Caveats:
  - Event study relies only on timing and sign restrictions; it does not control for other push-pull determinants of bilateral capital flows.
  - Empirical models have low R-squared values; much bilateral cross-border bank credit growth variation is unexplained.
  - Structural GFM simulations provide net general-equilibrium effects and cannot fully decompose partial-equilibrium credit leakage versus reallocation.
  - Case studies highlight circumvention and regulatory-perimeter widening as common responses.

### Structural simulation analysis (Global Macrofinancial Model, GFM)
- Model features:
  - Estimated New Keynesian DSGE model covering 40 countries; documented in Vitek (2017).
  - Includes nominal and real rigidities, bank and capital-market financial intermediation, and multiple international spillover channels.
- Spillover channels:
  - International trade linkages, international financial linkages (cross-border bank lending, nonfinancial private sector borrowing, portfolio debt and equity exposures), and commodity price linkages.
- Role of bilateral exposures:
  - Observed bilateral bank lending and nonfinancial private sector borrowing exposures average only 0.2 (value truncated in source) as narrow conduits for spillover transmission.
  - Many large bilateral exposures exist despite overall sparseness: 1.1 percent and 1.3 percent of them exceed 5 percent, respectively (referring to bilateral bank lending and nonfinancial private sector borrowing exposures).
- Simulation methodology:
  - Macrofinancial spillovers quantified using impulse responses to macroprudential policy shocks, identified as deviations from macroprudential policy rules.
  - Monetary policy responds endogenously; coordinated monetary and macroprudential policy measures are not considered.
  - Output and credit spillover coefficients defined as the ratio of the peak impulse response of output or bank credit in the recipient country to that in the source country.
  - Analysis covers 1560 directed country pairs.

### Instrument-specific GFM results and illustrative cases
- Countercyclical Capital Buffer (CCB):
  - CCB applies to total credit exposures and follows a rule responding inertially to bank credit growth, house and equity price growth.
  - Sweden → Finland illustrative case (CCB increase of one percentage point in Sweden):
    - Sweden: bank credit reduced by up to 0.14 percent.
    - Finland: bank credit reduced by up to 0.10 percent.
    - Implied output and credit spillover coefficients: 0.99 and 0.67, respectively.
  - Aggregate across 1560 directed pairs:
    - Average output spillover coefficient: 0.008.
    - Average credit spillover coefficient: 0.011.
    - Share of spillovers exceeding 0.05: 5.3 percent of output spillover coefficients; 4.9 percent of credit spillover coefficients.
    - Correlations: output with bilateral export exposures 0.56; credit with bilateral export exposures 0.36; output with bilateral nonfinancial private sector borrowing exposures 0.66; credit with bilateral nonfinancial private sector borrowing exposures 0.79.
  - Overall: CCB adjustments generate generally small global spillovers on average, but can produce large regionally concentrated spillovers for strongly linked pairs.
- Mortgage Loan-to-Value Limit (MLTVL):
  - MLTVL applies to borrowing by domestic developers and follows a rule responding inertially to mortgage debt growth and house price growth.
  - Germany → Czech Republic illustrative case (MLTVL reduction of one percentage point in Germany):
    - Germany: bank credit reduced by up to 0.53 percent; mortgage debt falls persistently.
    - Czech Republic: persistent output contraction of at most 0.05 percent.
    - Implied output spillover coefficient: 0.37.
  - Aggregate across 1560 directed pairs:
    - Average output spillover coefficient: 0.015.
    - Average credit spillover coefficient: 0.001.
    - Share of output spillovers exceeding 0.05: 6.8 percent.
    - Correlation between MLTVL output spillovers and bilateral export exposures: 0.91.
  - Overall: MLTVL adjustments generally generate small global spillovers on average, but some large, regionally concentrated spillovers occur for strongly linked partners.
- Corporate Loan-to-Value Limit (CLTVL):
  - CLTVL applies to borrowing by domestic firms from domestic and foreign banks and follows a rule responding inertially to nonfinancial corporate debt growth and equity price growth.
  - United States → Mexico illustrative case (CLTVL reduction of one percentage point in United States):
    - United States: bank credit reduced by up to 0.73 percent; nonfinancial corporate debt falls persistently.
    - Mexico: persistent hump-shaped export-driven output contraction; current account balance deteriorates.
    - Implied output spillover coefficient: 0.39.
  - Overall: CLTVL changes generate small average global spillovers but numerous large, regionally concentrated spillovers for strongly linked pairs.

### Cross-cutting quantitative findings
- 6.0 percent of output spillover coefficients exceed 0.05.
- 1.6 percent of credit spillover coefficients exceed 0.05.
- Correlations of estimated spillover coefficients with bilateral exposures:
  - Correlation of output spillover coefficients with bilateral export exposures: 0.87.
  - Correlation of credit spillover coefficients with bilateral export exposures: 0.45.
  - Correlation of output spillover coefficients with bilateral bank lending exposures: 0.31.
  - Correlation of credit spillover coefficients with bilateral bank lending exposures: 0.87.
- Aggregate GFM figures (across 1560 directed pairs) reported elsewhere include:
  - CCB average output spillover coefficient: 0.008.
  - MLTVL average output spillover coefficient: 0.015.
  - CCB average credit spillover coefficient: 0.011.
  - MLTVL average credit spillover coefficient: 0.001.
  - Example narrow-pair averages reported: 0.4 percent across the 1560 directed country pairs covered by the GFM (value context truncated in source).

### Network structure and case studies illustrating mechanisms
- GFM country coverage includes Argentina, Australia, Austria, Belgium, Brazil, Canada, Chile, China, Colombia, the Czech Republic, Denmark, Finland, France, Germany, Greece, India, Indonesia, Ireland, Israel, Italy, Japan, Korea, Malaysia, Mexico, the Netherlands, New Zealand, Norway, the Philippines, Poland, Portugal, Russia, Saudi Arabia, South Africa, Spain, Sweden, Switzerland, Thailand, Turkey, the United Kingdom, and the United States.
- Despite overall sparseness of the network, many large bilateral exposures exist; bilateral exposures are regionally concentrated and reflect relationship banking.
- CESEE case studies (pre-GFC and build-up):
  - Cross-border bank loans from advanced European source countries to Central Europe rose from about 10 to 30 percent of source GDP from 2002 to 2008.
  - Cross-border bank loans to Southeastern Europe rose from about 10 to 80 percent of source GDP from 2002 to 2008.
  - Tightening measures reduced credit growth and strengthened buffers, but circumvention occurred via cross-sectoral and cross-border credit leakage and regulatory-perimeter widening.
- Sweden → Baltic countries (pre-GFC):
  - Swedish bank loans to Baltic countries increased from about 15 percent of source GDP in 2003 to about 100 percent by 2007.
  - Baltic macroprudential tightening measures were partially ineffective at reducing credit growth because of high credit supply from Swedish banks and insufficient reciprocity, enabling regulatory arbitrage.

### Policy implications and conclusions
- Macroprudential policy taken in one jurisdiction can produce macrofinancial effects in others; when spillovers are significant, policymakers should consider them when addressing systemic risks.
- Empirical and simulation evidence together imply:
  - Sectoral and liquidity instruments can generate cross-border bank credit spillovers; capital instruments show less empirical evidence of cross-border spillovers in the event study and panel analyses presented.
  - Spillovers are stronger for tightenings, involve credit leakage and reallocation, and tend to be regionally concentrated.
  - Policymakers may need to consider reciprocity, widening regulatory perimeters, and harmonization to limit regulatory arbitrage and leakage.

*Source: IMF staff estimates and analysis (IMF Working Paper wp17170).*

### REFERENCES ___________________________________________________________42

### REFERENCES ___________________________________________________________42

### FIGURES

- 1. Cross-Border Bank Credit Leakage and Reallocation Effects _______________________9
- 2. Cross-Border Bank Credit Changes after Macroprudential Policy Tightening  _________13
- 3. Macroprudential Policy Tightening and Loosening Measures ______________________14
- 4. Bilateral Bank Lending and Nonfinancial Private Sector Borrowing Exposures ________24
- 5. Cluster Analysis of Bilateral Lending and Borrowing Networks ____________________25
- 6. Impulse Responses to a Unit CCB Increase ____________________________________27
- 7. Macrofinancial Spillovers from CCB Adjustments  ______________________________28
- 8. Impulse Responses to a Unit MLTVL Reduction ________________________________29
- 9. Macrofinancial Spillovers from MLTVL Adjustments  ___________________________30
- 10. Impulse Responses to a Unit CLTVL Reduction _______________________________31
- 11. Macrofinancial Spillovers from CLTVL Adjustments ___________________________32
- 12. Cross-Border Bank Loans from Advanced Europe to the CESEE Countries  _________35
- 13. Cross-Border Bank Loans from Sweden to the Baltic Countries ___________________37

### TABLES

- 1. Event Shares and Two-Sample Proportion Test Results, Full Sample ________________12
- 2. Event Shares and Two-Sample Proportion Test Results, Excluding the GFC __________15
- 3. Event Shares and Two-Sample Proportion Test Results, CESEE Countries ___________15
- 4. Hypothesis Test Results, Baseline Model Specification ___________________________20
- 5. Hypothesis Test Results, Full versus European Country Sample ____________________21
- 6. Hypothesis Test Results, Interactive Model Specification _________________________22
- 7. Panel Regression Variable Definitions ________________________________________38
- 8. Panel Regression Estimation Results, Baseline Model Specification  ________________39
- 9. Panel Regression Estimation Results, European Countries  ________________________40

*Source: wp17170 - REFERENCES ___________________________________________________________42*

### 10. Panel Regression Estimation Results, Interactive Model Specification ______________41

### 10. Panel Regression Estimation Results, Interactive Model Specification ______________41

### Major findings and conclusions
- Macroprudential policy measures targeted at the composition of banks’ funding (liquidity instruments) or the indebtedness of specific sectors (sectoral instruments) often have statistically significant cross-border bank credit spillover effects; measures targeted at the level or quality of banks’ capital buffers (capital instruments) generally do not.
- Empirical evidence of cross-border macroprudential policy spillovers is:
  - Stronger for tightening than for loosening measures.
  - Regionally concentrated (larger estimated spillovers in the European subsample).
- Event study and panel regression analyses:
  - Indicate statistically significant credit leakage and reallocation effects for liquidity and sectoral tightening measures.
  - Find little or no evidence of spillovers from capital instrument tightening, possibly because some capital measures are imposed on a consolidated basis.
- Structural model (GFM) simulations:
  - Indicate sectoral policy shocks are generally small worldwide but regionally concentrated and occasionally economically significant.
  - Indicate countercyclical capital buffer adjustments have the potential to generate sizeable regional spillovers.
- Estimated model fit:
  - The baseline panel regression explains a small proportion of variation in directional bilateral cross-border bank credit growth, with R-squared values of about 0.025 for the full sample and 0.050 for the European country subsample.

### Data, coverage, and indices
- Panel coverage:
  - 64 countries over the period 2000Q1 to 2015Q1.
  - Panel dimensions reported: 64N countries and 61T time periods, implying a total of 3,904NT observations.
- Bank credit measures:
  - Domestic bank credit, cross-border bank lending and borrowing, and directional bilateral cross-border bank credit from the BIS International Banking Statistics.
  - Locational cross-border bank credit data are used (based on residency principle).
  - Bank assets in the form of loans are the focus; securities and other claims excluded.
- Macroprudential policy database:
  - New database covering 11 frequently used instruments over 2000Q1 to 2015Q1.
  - Functional classification of the 11 instruments:
    - Capital instruments: time-varying capital requirements, dynamic loan-loss provisions, profit distribution restrictions, countercyclical capital buffers (countercyclical capital buffer adjustments analyzed by structural methods only).
    - Liquidity instruments: loan-to-deposit ratio limits, maturity mismatch limits, net open foreign exchange position limits, reserve requirements.
    - Sectoral instruments: loan-to-value ratio limits, debt-service-to-income ratio limits, debt-to-income or loan-to-income ratio limits, quantitative loan restrictions.
  - Indices:
    - Country-specific, time-varying indicator indices that equal one if at least one instrument in the given category is tightened or loosened at least once during the quarter, and zero otherwise.
    - Indices capture direction but not size of instrument adjustments.
- Special samples and periods:
  - Global Financial Crisis (GFC) period defined as 2008Q1 to 2012Q4.
  - Excluding the GFC period (2008Q1–2012Q4) yields broadly consistent event study results; exclusion almost halves the number of tightening/loosening events under consideration.
  - CESEE (Central, Eastern and Southeastern European) country subsample: Bulgaria, Croatia, the Czech Republic, Hungary, Poland, Romania, and the Slovak Republic.

### Event study methodology and key empirical signals
- Identification:
  - Nonparametric event study uses only timing and sign restrictions.
  - Persistent spillover events require sign restrictions to be satisfied for at least two consecutive quarters on a year-over-year basis.
- Definitions:
  - Credit leakage: domestic tightening (loosening) associated with increased (decreased) borrowing from foreign banks.
  - Credit reallocation: domestic tightening (loosening) associated with increased (decreased) cross-border lending by domestic banks to foreign borrowers.
- Event study results (summary):
  - Tightening measures using liquidity or sectoral instruments are associated with:
    - Lower domestic bank credit growth.
    - Higher growth in cross-border borrowing from foreign banks.
    - Higher growth in domestic banks’ cross-border lending to foreign borrowers.
  - Little or no evidence of spillovers from tightening measures using capital instruments.
  - Little or no evidence of cross-border spillovers from loosening measures overall (with some exceptions when excluding GFC).
- Magnitude signals from event study medians:
  - On average, cross-border bank borrowing or lending grew 2 to 4 percentage points faster within two quarters of macroprudential policy tightening using liquidity or sectoral instruments (median changes in quarterly year-over-year growth rates).
  - Cross-border bank borrowing or lending grew 2 to 5 percentage points slower within two quarters after macroprudential policy tightening using capital instruments (median changes).

### Panel regression model specification, estimation, and inference
- Theoretical framing:
  - Bilateral cross-border bank credit modeled via a gravity-type relationship: proportional to recipient demand and source supply, inversely proportional to financial distance.
  - Recipient demand functions and source supply functions specified; financial distance depends on financial market development, capital flow restrictions, exchange rate flexibility, financial stress, and macroprudential policy indices.
- Reduced-form estimation equation (growth rate of directional bilateral cross-border bank credit) includes:
  - Differences in recipient and source domestic credit growth, nominal output growth, nominal bank lending rate changes, real house price index growth.
  - Differences in financial market development indices, capital flow restriction intensity indices, exchange rate flexibility indices.
  - Cumulated financial crisis indicator and cumulated instrument-specific macroprudential tightening/loosening indices.
  - Pairwise fixed effects and time fixed effects included; likelihood ratio tests support their inclusion.
- Estimation details:
  - Annual frequency over 2000–2014 (dataset converted to annual).
  - Dependent variable winsorized at the 5 percent level to remove outliers.
  - Instrumental variables used to correct for endogeneity where required:
    - Domestic bank credit growth in recipient and source countries treated as endogenous.
    - Instruments: growth rate of nominal output, change in nominal bank lending interest rate, growth rate of real house price index (in recipient and source countries).
    - Validity checks performed: Davidson and MacKinnon exogeneity test, Kleibergen and Paap underidentification test, Hansen overidentification test.
  - Dynamic specification rejected due to overidentification test failures for reasonable lag orders; static specification favored.
- Panel regression inference (hypothesis test summaries):
  - Strong evidence (1 percent level) of credit leakage effects from tightening in recipient countries using liquidity instruments (higher growth in cross-border borrowing from foreign banks).
  - Moderate evidence (5 percent level) of credit reallocation effects from tightening of sectoral instruments in source countries (higher growth in cross-border lending by domestic banks to foreign borrowers).
  - Moderate evidence that loosening measures in source countries using liquidity and sectoral instruments are associated with lower growth in domestic bank lending to foreign borrowers.
  - Little or no evidence of leakage or reallocation effects from tightening or loosening measures using capital instruments.
- European subsample results:
  - Broadly consistent with full sample but quantitatively larger: estimated credit leakage and reallocation effects from liquidity and sectoral tightening measures are about 50 and 100 percent larger, respectively, relative to full sample estimates—indicative of regional concentration of spillovers.
  - Additional evidence in European subsample: weak evidence of leakage effects from sectoral tightening; moderate evidence of reallocation effects from liquidity tightening in source countries.
- Interactive model specification:
  - Macroprudential tightening/loosening indices interacted linearly with capital flow restriction intensity indices (to capture dependence on capital account openness).
  - Findings:
    - Strong evidence that leakage effects from liquidity tightening are larger in recipient countries with more open capital accounts.
    - Weak evidence that reallocation effects from sectoral tightening are larger in source countries with more open capital accounts.
    - Strong evidence that reallocation effects from sectoral loosening are larger in source countries with more open capital accounts.
    - Little or no evidence that capital instrument spillovers are amplified by more open capital accounts.
- Model fit caveat:
  - Low explanatory power: baseline model R-squared about 0.025 (full sample) and 0.050 (European subsample), despite statistically significant coefficients for key hypotheses.

### Robustness checks and caveats
- Robustness exercises:
  - Excluding the GFC period (2008Q1–2012Q4).
  - Focusing on CESEE subsample.
  - Interactive specification with capital account openness.
- Event study caveat:
  - Event study relies only on timing and sign restrictions; it does not control for other push-pull determinants of bilateral capital flows.
- Panel regression caveats:
  - Low R-squared values imply most variation in bilateral cross-border bank credit growth remains unexplained by the model.
  - Empirical inability to fully separate credit leakage versus reallocation in structural equilibrium models; structural GFM provides net general-equilibrium effects rather than partial-equilibrium leakage vs reallocation decomposition.
- Case-study context:
  - CESEE case studies highlight circumvention and regulatory-perimeter widening: prior to the GFC, tightening measures in some CESEE countries coincided with accelerated borrowing from foreign banks and credit intermediation migration to nonbanks; authorities later widened the regulatory perimeter to limit arbitrage.

### Structural simulation analysis (Global Macrofinancial Model, GFM)
- Model features:
  - Estimated New Keynesian DSGE model (GFM documented in Vitek (2017)).
  - Covers 40 countries.
  - Includes nominal and real rigidities, bank and capital-market financial intermediation, and multiple international spillover channels.
- Spillover channels in the structural model:
  - International trade linkages: export exposures transmit demand and exchange-rate shifts induced by macroprudential policy changes.
  - International financial linkages: cross-border bank lending, nonfinancial private sector borrowing, portfolio debt and equity exposures, and contagion effects; most important conduits depend on whether measures are imposed on banks or on firms.
  - Commodity price linkages: matter when source countries are large enough to alter world commodity prices, affecting recipients via terms-of-trade shifts.
- Role of bilateral exposures:
  - Observed bilateral bank lending and nonfinancial private sector borrowing exposures are generally small relative to total lending and borrowing: they average only 0.2 (value truncated in source) as narrow conduits for spillover transmission.
  - Structural simulations estimate economic significance of full transmission mechanisms, but cannot decompose partial-equilibrium leakage versus reallocation effects.

*Source: IMF staff estimates and analysis, “10. Panel Regression Estimation Results, Interactive Model Specification,” IMF Working Paper wp17170.*

### 0.4 percent across the 1560 directed country pairs covered by the GFM, respectively.

### wp17170 - 0.4 percent across the 1560 directed country pairs covered by the GFM, respectively.

### Network structure and bilateral exposures
- The global financial network (GFM) covers the following countries: Argentina, Australia, Austria, Belgium, Brazil, Canada, Chile, China, Colombia, the Czech Republic, Denmark, Finland, France, Germany, Greece, India, Indonesia, Ireland, Israel, Italy, Japan, Korea, Malaysia, Mexico, the Netherlands, New Zealand, Norway, the Philippines, Poland, Portugal, Russia, Saudi Arabia, South Africa, Spain, Sweden, Switzerland, Thailand, Turkey, the United Kingdom, and the United States.
- Despite overall sparseness, there are many large bilateral exposures:
  - 1.1 percent and 1.3 percent of them exceed 5 percent, respectively (referring to bilateral bank lending and nonfinancial private sector borrowing exposures).
- Bilateral bank lending and nonfinancial private sector borrowing exposures are regionally concentrated, reflecting relationship banking:
  - On the lending side, the banking systems of the United Kingdom and United States have high network centrality; France, Spain and Sweden act as regional gatekeepers.
  - On the borrowing side, the nonfinancial private sectors of Austria, Belgium, France, Spain and Sweden have multiple strong regional network connections.

### Simulation methodology and definitions
- Macrofinancial spillovers are quantified using impulse responses to macroprudential policy shocks, identified as deviations from macroprudential policy rules.
- Monetary policy is allowed to respond endogenously to macroprudential policy shocks; coordinated monetary and macroprudential policy measures are not considered.
- Output and credit spillover coefficients are defined as the ratio of the peak impulse response of output or bank credit in the recipient country to that in the source country.
- The analysis covers 1560 directed country pairs.

### Countercyclical Capital Buffer (CCB) — transmission and magnitudes
- In the GFM, the CCB applies to total credit exposures (domestic mortgage and corporate loans, and foreign corporate loans) and follows a rule that responds inertially to bank credit growth, house and equity price growth.
- Sweden → Finland illustrative case (CCB increase of one percentage point in Sweden):
  - Sweden: bank credit reduced by up to 0.14 percent.
  - Finland: bank credit reduced by up to 0.10 percent.
  - Implied output and credit spillover coefficients: 0.99 and 0.67, respectively.
  - Channels: increased spreads on mortgage and corporate loan rates, retained earnings accumulation to satisfy higher capital ratio, output contraction in source country, monetary easing, currency depreciation, fiscal balance deterioration (due to fall in nominal output), current account improvement (import compression).
- Aggregate estimated spillovers from CCB adjustments across 1560 directed pairs:
  - Average output spillover coefficient: 0.008.
  - Average credit spillover coefficient: 0.011.
  - Share of spillovers exceeding 0.05: 5.3 percent of output spillover coefficients; 4.9 percent of credit spillover coefficients.
  - Correlations:
    - Output spillover coefficients with bilateral export exposures: 0.56.
    - Credit spillover coefficients with bilateral export exposures: 0.36.
    - Output spillover coefficients with bilateral nonfinancial private sector borrowing exposures: 0.66.
    - Credit spillover coefficients with bilateral nonfinancial private sector borrowing exposures: 0.79.
- Overall assessment: CCB adjustments generate generally small global spillovers on average, but can produce large regionally concentrated spillovers for strongly linked country pairs.

### Mortgage Loan-to-Value Limit (MLTVL) — transmission and magnitudes
- In the GFM, the MLTVL applies to borrowing by domestic developers from domestic banks and follows a rule that responds inertially to mortgage debt growth and house price growth.
- Germany → Czech Republic illustrative case (MLTVL reduction of one percentage point in Germany):
  - Germany: bank credit reduced by up to 0.53 percent; mortgage debt falls persistently.
  - Czech Republic: persistent output contraction of at most 0.05 percent.
  - Implied output spillover coefficient: 0.37.
  - Channels: binding quantitative restriction reducing mortgage debt and bank credit, output contraction in source country, monetary easing in union or recipient, currency depreciation, fiscal deterioration, current account effects via export reductions.
- Aggregate estimated spillovers from MLTVL adjustments across 1560 directed pairs:
  - Average output spillover coefficient: 0.015.
  - Average credit spillover coefficient: 0.001.
  - Share of output spillovers exceeding 0.05: 6.8 percent.
  - Correlation between estimated output spillover coefficients and bilateral export exposures: 0.91.
- Overall assessment: MLTVL adjustments generally generate small global spillovers on average, but some large, regionally concentrated spillovers occur for strongly linked partners.

### Corporate Loan-to-Value Limit (CLTVL) — transmission and magnitudes
- In the GFM, the CLTVL applies to borrowing by domestic firms from domestic and foreign banks and follows a rule that responds inertially to nonfinancial corporate debt growth and equity price growth.
- United States → Mexico illustrative case (CLTVL reduction of one percentage point in United States):
  - United States: bank credit reduced by up to 0.73 percent; nonfinancial corporate debt falls persistently.
  - Mexico: persistent hump-shaped export-driven output contraction; current account balance deteriorates (lower exports to United States and lower commodity prices).
  - Implied output spillover coefficient: 0.39.
- Overall assessment: CLTVL changes generate small average global spillovers but numerous large, regionally concentrated spillovers for strongly linked country pairs.

### Cross-cutting findings and implications
- Macroprudential policy measures in one country can transmit sizable macrofinancial spillovers to other countries when bilateral financial and trade linkages are strong and regionally concentrated.
- On average, estimated spillover coefficients are small (examples: CCB output 0.008 average; MLTVL output 0.015 average), but a nontrivial fraction of directed pairs exhibit large spillovers (e.g., 5.3 percent of CCB output spillovers > 0.05; 6.8 percent of MLTVL output spillovers > 0.05).
- The magnitude of spillovers correlates strongly with bilateral exposures:
  - Export exposures are highly correlated with MLTVL output spillovers (0.91).
  - Nonfinancial private sector borrowing exposures are strongly correlated with CCB credit spillovers (0.79).

*Source: IMF staff estimates derived from the Global Financial Model (GFM) and BIS International Banking Statistics, as presented in the source PDF.*

### 6.0 percent of our output and 1.6 percent of our credit spillover coefficients exceed 0.05. The

### wp17170 - 6.0 percent of our output and 1.6 percent of our credit spillover coefficients exceed 0.05. The

### Key quantitative findings
- 6.0 percent of our output spillover coefficients exceed 0.05.
- 1.6 percent of our credit spillover coefficients exceed 0.05.
- Correlations of estimated spillover coefficients with bilateral exposures:
  - Correlation of output spillover coefficients with bilateral export exposures: 0.87.
  - Correlation of credit spillover coefficients with bilateral export exposures: 0.45.
  - Correlation of output spillover coefficients with bilateral bank lending exposures: 0.31.
  - Correlation of credit spillover coefficients with bilateral bank lending exposures: 0.87.

### Simulation analysis results (macroprudential instruments)
- General result:
  - Macrofinancial spillovers from macroprudential policy measures using capital and sectoral instruments are generally small worldwide, but adjustments in some countries can have large macrofinancial spillovers to others.
- Specific instrument findings:
  - CCB (countercyclical capital buffer) adjustments:
    - Output and bank credit spillovers from CCB adjustments can be economically significant for recipient countries having high export or nonfinancial private sector borrowing exposures to source countries.
    - Simulation indicates CCB adjustments have the potential to generate sizeable regional macrofinancial spillovers.
    - Panel data set does not cover widespread recent CCB use; future empirical analyses may reveal regionally concentrated cross-border bank credit spillovers from CCB adjustments as observations accumulate.
  - MLTVL (mortgage loan-to-value limit) adjustments:
    - Output spillovers from MLTVL adjustments can be economically significant for recipient countries having high export exposures to source countries.
  - CLTVL (corporate loan-to-value limit) adjustments:
    - Output and bank credit spillovers from CLTVL adjustments can be economically significant for recipient countries having high export or bank lending exposures to source countries.
- Geographic and network context:
  - Most country pairs in sample are weakly connected by trade and financial linkages, but important regionally concentrated linkages exist, explaining concentrated large spillovers.

### Empirical evidence (event study and panel regressions)
- Methods used:
  - Nonparametric event study analysis.
  - Reduced-form panel regression analysis.
  - Structural model based simulation analysis.
- Main empirical findings:
  - Evidence of cross-border bank credit spillovers from sectoral and liquidity based macroprudential policy measures.
  - No evidence of cross-border bank credit spillovers from capital measures in event study and panel regression analyses.
  - Empirical evidence is stronger for tightening than for loosening measures.
  - Spillovers are distributed across credit leakage and reallocation effects.
  - Spillovers are generally regionally concentrated.
- Consistency:
  - Simulation results are consistent with empirical results: sectoral instrument adjustments (MLTVL and CLTVL) are generally small worldwide but can be large among countries with high export or bank lending exposures.

### Case studies (illustrating cross-border bank credit spillovers undermining macroprudential effectiveness)
- Central and Southeastern Europe (CESEE), build-up to the GFC:
  - Cross-border bank loans from advanced European source countries (Austria, Belgium, Italy, Switzerland) to Central Europe rose from about 10 to 30 percent of source GDP from 2002 to 2008.
  - Cross-border bank loans to Southeastern Europe rose from about 10 to 80 percent of source GDP from 2002 to 2008.
  - Systemic risks arose mainly from currency mismatches on bank, household and nonfinancial corporate balance sheets.
  - Macroprudential tightening measures taken in mid-2000s included:
    - Raising risk weights on foreign currency denominated bank loans (Croatia, Poland, Serbia).
    - Tightening lending standards and raising loan-loss provisions (Romania).
    - Imposing marginal reserve requirements on excessive credit growth and external borrowing (Bulgaria, Croatia).
    - Imposing loan-to-value and debt-service-to-income ratio limits (Bulgaria, Croatia).
  - Outcomes:
    - Measures were partially effective at reducing credit growth and strengthened capital and liquidity buffers.
    - Circumvention occurred via cross-sectoral and cross-border credit leakage: migration to domestic nonbanks and displacement by cross-border borrowing from foreign banks.
    - Example: Borrowing from foreign banks accelerated in Bulgaria and Croatia; in Poland borrowing was channeled through foreign branches not subject to domestic macroprudential policy.
    - Authorities responded by widening regulatory perimeter and harmonizing financial regulation.
  - Additional effect:
    - Cross-border reallocation effects: tightening in Advanced Europe targeted internationally active banks (higher capital, restrictions on non-lending activities), prompting European banks to search for yield and weaken lending standards to CESEE (Ongena, Popov and Udell (2013) referenced).

- Sweden and the Baltic Countries, pre-GFC:
  - Swedish bank loans to Baltic countries increased from about 15 percent of source GDP in 2003 to about 100 percent by 2007.
  - Systemic risks arose from currency mismatches and overvalued housing markets.
  - Baltic macroprudential tightening measures included:
    - Raising risk weights on mortgage loans.
    - Raising reserve requirements.
    - Imposing loan-to-value and debt-service-to-income ratio limits on households (Latvia).
    - Reducing tax deductibility of mortgage interest payments (Estonia).
  - Outcomes:
    - Measures were ineffective at reducing credit growth but strengthened banks’ capital and liquidity buffers.
    - Large cross-border bank credit spillovers and credit leakage were major contributors to ineffectiveness:
      - High credit supply from Swedish banks exploiting profitable lending opportunities given low-cost global wholesale funding.
      - Circumvention through cross-border borrowing from Swedish banks via local subsidiaries.
      - Insufficient reciprocity in Sweden of Baltic macroprudential tightening measures allowed regulatory arbitrage by local bank branches.

### Panel regression analysis (variables and estimation notes)
- Dependent variable:
  - D_lnTL_W: Growth rate of directional bilateral cross-border bank credit, winsorized at 5 percent level.
- Selected explanatory variables (definitions preserved):
  - creditgr_recipient: Growth rate of domestic bank credit in recipient country.
  - creditgr_source: Growth rate of domestic bank credit in source country.
  - D.fdel_recipient: Change in financial market development index for recipient country.
  - D.fdel_source: Change in financial market development index for source country.
  - D.schkin_recipient: Schindler index of capital inflow controls in recipient country (higher value indicates more/tighter controls).
  - D.schkout_source: Schindler index of capital outflow controls in source country (higher value indicates more/tighter controls).
  - D.erri_recipient: Change in exchange rate regime in recipient country (higher value indicates more flexible regime).
  - D.erri_source: Change in exchange rate regime in source country (higher value indicates more flexible regime).
  - crisis_recipient: Dummy = 1 if banking/currency/sovereign crisis in recipient country, 0 otherwise.
  - crisis_source: Dummy = 1 if banking/currency/sovereign crisis in source country, 0 otherwise.
  - tallsum_recipient / tallsum_source: Tightening of any macroprudential instrument in recipient/source country.
  - lallsum_recipient / lallsum_source: Loosening of any macroprudential instrument in recipient/source country.
  - tcapital_*, lcapital_*: Tightening/loosening of macroprudential capital instrument in recipient/source country.
  - tliquidity_*, lliquidity_*: Tightening/loosening of macroprudential liquidity instrument in recipient/source country.
  - tsectoral_*, lsectoral_*: Tightening/loosening of macroprudential sectoral instrument in recipient/source country.
  - tborrower_*, lborrower_*: Tightening/loosening of macroprudential policy on borrowers in recipient/source country.
  - tlender_*, llender_*: Tightening/loosening of macroprudential policy on lenders in recipient/source country.
- Instrumental variables:
  - GDPGR_recipient: Growth rate of nominal gross domestic product in recipient country.
  - GDPGR_source: Growth rate of nominal gross domestic product in source country.
  - int_recipient / int_source: Change in nominal bank lending interest rate in recipient/source country.
  - HPGR_recipient / HPGR_source: Growth rate of real house price index in recipient/source country.
- Estimation notes:
  - Panel regression estimation results are reported in Table 8 (baseline), Table 9 (European countries), and Table 10 (interactive model specification).
  - Significance indicators: ***, **, and * denote significance at the 1, 5 and 10 percent levels on the basis of robust standard errors.

### Conclusion and policy implications
- Macroprudential policy is widely used in advanced and emerging market economies; such measures taken in one jurisdiction can produce macrofinancial effects in others.
- If spillovers are significant, policymakers should consider them when addressing systemic risks.
- Empirical and simulation evidence together imply:
  - Sectoral and liquidity instruments can generate cross-border bank credit spillovers; capital instruments show less empirical evidence of cross-border spillovers in the event study and panel analyses presented.
  - Spillovers are stronger for tightenings, involve credit leakage and reallocation, and tend to be regionally concentrated.
  - Policymakers may need to consider reciprocity, widening regulatory perimeters, and harmonization to limit regulatory arbitrage and leakage.

*Source: IMF staff estimates (content from the provided IMF Working Paper excerpt).*

### REFERENCES

### REFERENCES

### Works cited

- Aiyar, S., C. Calomiris, J. Hooley, Y. Korniyenko and T. Wieladek, 2014, “The International Transmission of Bank Capital Requirements: Evidence from the UK”, Journal of Financial Economics, Vol. 113, pp. 368−382.  
- Akinci, O. and J. Olmstead-Rumsey, 2015, “How Effective are Macroprudential Policies? An Empirical Investigation”, Board of Governors of the Federal Reserve System International Finance Discussion Paper, 1136.  
- Bakker, B. and A. Gulde, 2010, “The Credit Boom in the EU New Member States: Bad Luck or Bad Policies?”, International Monetary Fund Working Paper, 130.  
- Bengui, J., 2012, “Macro-Prudential Policy Coordination and Global Regulatory Spillovers”, Unpublished Manuscript.  
- Bergstrand, J., 1985, “The Gravity Equation in International Trade: Some Microeconomic Foundations and Empirical Evidence”, The Review of Economics and Statistics, Vol. 67, pp. 474−481.  
- Buch, C. and L. Goldberg, 2016, “Cross-Border Prudential Policy Spillovers: How Much? How Important? Evidence from the International Banking Research Network”, National Bureau of Economic Research Working Paper, 22874.  
- Cerutti, E., S. Claessens and L. Laeven, 2015, “The Use and Effectiveness of Macroprudential Policies: New Evidence”, International Monetary Fund Working Paper, 61.  
- Cizel, J., J. Frost, A. Houben and P. Wierts, 2016, “Effective Macroprudential Policy: Cross-Sector Substitution from Price and Quantity Measures,” International Monetary Fund Working Paper, 94.  
- Davidson, R. and J. MacKinnon, 1993, Estimation and Inference in Econometrics, Oxford University Press.  
- Deardorff, A., 1998, “Determinants of Bilateral Trade: Does Gravity Work in a Neoclassical World?”, The Regionalization of the World Economy, edited by J. Frankel, University of Chicago Press.  
- Fahr, S. and D. Zochowski, 2015, “A Framework for Analysing and Assessing Cross-Border Spillovers from Macroprudential Policies”, European Central Bank Financial Stability Review, May.  
- Fernández, A., M. Klein, A. Rebucci, M. Schindler and M. Uribe, 2015, “Capital Control Measures: A New Dataset”, International Monetary Fund Working Paper, 80.  
- Hansen, L., 1982, “Large Sample Properties of Generalized Method of Moments Estimators”, Econometrica, Vol. 50, pp. 1029-1054.  
- Herrmann, S. and D. Mihaljek, 2013, “The Determinants of Cross-Border Bank Flows to Emerging Markets: New Empirical Evidence on the Spread of Financial Crises”, Economics of Transition, Vol. 21, pp. 479-508.  
- Houston, J., C. Lin and Y. Ma, 2012, “Regulatory Arbitrage and International Bank Flows”, The Journal of Finance, Vol. 67, pp. 1845−1895.  
- International Monetary Fund, 2014a, “Staff Guidance Note on Macroprudential Policy”, International Monetary Fund Staff Guidance Note.  
- International Monetary Fund, 2014b, “Staff Guidance Note on Macroprudential Policy―Detailed Guidance on Instruments”, International Monetary Fund Staff Guidance Note.  
- International Monetary Fund, Financial Stability Board, and Bank for International Settlements, 2016, “Elements of Effective Macroprudential Policies: Lessons from International Experience,” Report to the G20.  
- Kleibergen, F. and R. Paap, 2006, “Generalized Reduced Rank Tests using the Singular Value Decomposition”, Journal of Econometrics, Vol. 127, pp. 97−126.  
- Laeven, L. and F. Valencia, 2012, “Systemic Banking Crises Database: An Update”, International Monetary Fund Working Paper, 163.  
- Lim, C., F. Columba, A. Costa, P. Kongsamut, A. Otani, M. Saiyid, T. Wezel and X. Wu, 2011, “Macroprudential Policy: What Instruments and How to Use Them? Lessons from Country Experiences”, International Monetary Fund Working Paper, 238.  
- Lim, C., I. Krznar, F. Lipinsky, A. Otani and X. Wu, 2013, “The Macroprudential Framework: Policy Responsiveness and Institutional Arrangements”, International Monetary Fund Working Paper, 166.  
- Ongena, S., A. Popov and G. Udell, 2013, “When the Cat’s Away the Mice will Play: Does Regulation at Home Affect Bank Risk-Taking Abroad?”, Journal of Financial Economics, Vol. 108, pp. 727−750.  
- Portes, R. and H. Rey, 2005, “The Determinants of Cross-Border Equity Flows”, Journal of International Economics, Vol. 65, pp. 269−296.  
- Reinhardt, D. and R. Sowerbutts, 2015, “Regulatory Arbitrage in Action: Evidence from Banking Flows and Macroprudential Policy”, Bank of England Working Paper, 546.  
- Sahay, R., M. Cihák, P. N’Diaye, A. Barajas, R. Bi, D. Ayala, Y. Gao, A. Kyobe, L. Nguyen, C. Saborowski, K. Svirydzenka and S. Yousefi, 2015, “Rethinking Financial Deepening: Stability and Growth in Emerging Markets”, International Monetary Fund Staff Discussion Note, 8.  
- Sutt, A., H. Korju and K. Siibak, 2011, “The Role of Macro-Prudential Policies in the Boom and Adjustment Phase of the Credit Cycle in Estonia”, World Bank Policy Research Working Paper, 5835.  
- Viñals, J. and E. Nier, 2014, “Collective Action Problems in Macroprudential Policy and the Need for International Coordination”, Banque de France Financial Stability Review, Vol. 18, pp. 39−46.  
- Vitek, F. (2017), “Policy, Risk and Spillover Analysis in the World Economy: A Panel Dynamic Stochastic General Equilibrium Approach”, International Monetary Fund Working Paper, 89.

*Source: wp17170 - REFERENCES*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp17170.pdf_