## _wp14188

## Source details

**Canonical URL:** [_wp14188](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14188.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14188.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14188.pdf.json)

---

### Introduction and research questions
- Investigates whether adoption of capital controls or prudential measures (collectively, capital account restrictions, CARs) on outflows by source countries and on inflows by recipient countries can moderate large and potentially disruptive cross-border bank flows.
- Uses bilateral cross-border bank flow data to assess the joint effectiveness and interaction of CARs at both ends.

### Data, sample, and measurement
- Bilateral data on cross-border bank flows from 31 major source to 76 major recipient countries over 1995–2012.
- Flows: exchange rate-adjusted changes in gross international financial claims of resident banks on bank and nonbank sectors of recipient countries (BIS Locational Banking Statistics by Residence); data cover over 90 percent of international assets of domestic banking institutions.
- CARs drawn from IMF’s AREAER and the OECD Code of Liberalization of Capital Movements.
- Prudential measures captured as binary variables (1 = restriction present, 0 = otherwise).
- Capital control measures constructed following Schindler (2009); asset-class averages take values 0, 0.5, or 1 for bond/equity controls and 0 or 1 for direct investment and financial credit where disaggregation is limited.
- Overall restrictiveness indices: averages over asset classes, range between 0 and 1.
- Lagging of country-specific variables by one period; cluster standard errors at country-pair level.

### Empirical strategy and robustness
- Gravity-type panel regressions:
  - Equation (1): source-side outflow CARs and controls.
  - Equation (2): recipient-side inflow CARs and controls.
  - Equation (3): simultaneous inclusion of source outflow and recipient inflow CARs.
- Controls: (log) real GDP, real GDP per capita, real interest rates, real GDP growth, current account balance (percent of GDP), exchange rate regime (peg = 1), global VIX, commodity prices, time and country-pair fixed effects.
- Robustness checks: OLS, CPFE/TE, alternative control sets, alternate dependent variables (flows and stocks), alternate CAR definitions, IV-2SLS, lag structures (1-year lag and 5-year lagged averages).
- IV instruments: Monetary freedom (Heritage Foundation Economic Freedom Index) and presence of a left-wing government (Beck et al., 2001).

### Stylized statistics (selected)
- Total bank asset flows from reporting advanced source countries to advanced recipient countries:
  - about USD 435 billion in 1995;
  - almost USD 4 trillion in 2007;
  - about USD 280 billion in 2010.
- On the eve of the crisis in 2007, bank inflows constituted:
  - about 50 percent of gross inflows to advanced countries;
  - about 40 percent of gross inflows to EMs.
- Proportion of EMs with restrictions on lending locally in FX:
  - about 50 percent in 2001;
  - 57 percent in 2007;
  - 70 percent in 2012.
- Proportion of EMs with restrictions on purchase of locally issued securities denominated in FX:
  - 42 percent in 2001;
  - 46 percent in 2007;
  - 49 percent in 2012.
- Benchmark sample and estimation context:
  - Observations: 22,257.
  - No. of source (recipient) countries: 31 (76).
  - No. of country-pairs: 1,943.
  - R2 (within CPFE): 0.05 reported in many CPFE/TE specifications.

### Main empirical findings — Source country (outflow) CARs
- Overall restrictiveness on capital outflows:
  - OLS: moving from the 25th to the 75th percentile associated with about 70 percent lower outflows.
  - CPFE/TE: same movement implies about 77 percent lower outflows.
- Disaggregated outflow controls (bond, equity, direct investment, financial credit):
  - Moving from the bottom to the top quartile reduces flows by some 50-100 percent.
- Prudential measure — restrictions on lending to nonresidents:
  - Associated with about 100 percent reduction in bank asset flows (statistically significant).
- Source country real interest rate:
  - A 100 basis points increase reduces bank outflows by about 20 percent (CPFE/TE).
- Selected regression coefficients (Table 2 highlights):
  - Capital outflow controls (i): -2.442*** (col. (1)) and -2.978*** (col. (2)).
  - Bond outflow controls (i): -1.446* (col. (3)).
  - Equity outflow controls (i): -2.316*** (col. (4)).
  - Direct investment outflow controls (i): -4.019*** (col. (5)).
  - Financial credit outflow controls (i): -1.438*** (col. (6)).
  - Lending to nonresidents (i): -5.505*** (col. (7)).
  - Log (VIX): -7.933***.

### Main empirical findings — Recipient country (inflow) CARs
- Overall capital inflow controls (CPFE/TE):
  - Moving from the 25th to the 75th percentile associated with about 50 percent lower bank asset flows (significant at 5 percent).
- Bond inflow controls:
  - Moving from the 25th to the 75th percentile lowers bank inflows by about 64 percent.
- Equity and direct investment inflow controls:
  - Statistically insignificant in baseline CPFE/TE results.
- Financial-credit inflow controls:
  - Marginally insignificant (p-value = 0.11).
- FX-related prudential measures in recipients:
  - Restrictions on lending locally in FX and open FX position limits: inflows about 70-80 percent lower when present.
- Selected regression coefficients (Table 3 highlights):
  - Bond inflow controls (j): -2.023*** (col. (3)).
  - Lending locally in FX (j): -1.792*** (col. (7)).
  - Open FX position limits (j): -1.110* (col. (9)).
  - Log (VIX): -7.716***.

### Joint implementation and interaction of CARs
- Simultaneous inclusion of source outflow and recipient inflow CARs yields coefficients similar in magnitude and significance to individual estimates.
- Combined effects are not fully additive (log specification). Example using reported estimates:
  - Source overall outflow index: 25th to 75th percentile → about 80 percent reduction (holding recipient constant).
  - Recipient bond inflow index: 25th to 75th percentile → about 50 percent reduction (holding source constant).
  - Joint impact of both measures → 63 percent reduction.
- Effectiveness conditional on openness of the other end:
  - Outflow measures have larger impact when recipient country is more open (below 75th percentile on its inflow index).
  - Inflow measures have larger impact when source country is more open (below 75th percentile on its outflow index).
  - When the other side is already relatively closed, incremental impact is smaller and often statistically insignificant.
- Table 4 highlights:
  - Capital outflow controls index (i): -2.999*** (col. (1)).
  - Capital inflow controls index (j): -2.027** (col. (2)).
  - Direct investment outflow controls index (i): -4.020***.
  - Financial credit outflow controls index (i): -1.443***.
  - Lending to nonresidents (i): -5.489***.
  - Lending locally in FX (j): -1.801***.
  - Open FX position limits (j): -1.116*.

### Counterfactual scenarios (Table 4 col. [6] estimates)
- Uses simulations for 2007 (pre-GFC) with financial credit outflow controls (source) and restrictions on lending locally in FX (recipient):
  1. All source countries impose financial credit outflow controls; other variables at actual 2007 levels.
  2. All recipient countries impose restrictions on lending locally in FX; other variables at actual 2007 levels.
  3. Both all source and all recipient countries impose the respective measures; other variables at actual 2007 levels.
- Selected regional simulated changes in predicted 2007 flows:
  - Emerging Europe:
    - Scenario 1: about 20 percent lower flows.
    - Scenario 2: about 80 percent lower flows.
    - Scenario 3: about 85 percent lower flows.
  - Eurozone peripheral countries (Greece, Ireland, Italy, Portugal and Spain):
    - Scenario 3: flows might have been almost 95 percent lower.
  - Latin America:
    - Scenario 1: smaller effect because some source countries already had controls.
    - Scenario 2: flows lowered by about 80 percent.
  - Asia:
    - Scenario 1: source country action reduces flows by over 30 percent.
    - Scenario 2: smaller effect since many recipient Asian countries already had FX lending restrictions in 2007.
  - Post-GFC surge control:
    - Action by all source and recipient countries would have lowered inflows to Asian and Latin American EMs by about 10 and 90 percent, respectively.

### Spillover effects
- Evidence of capital flow deflection:
  - Raising average inflow controls in regional or economically similar neighbors increases bank flows to a recipient country.
  - Raising average inflow controls in regional neighbors from the median to the 75th percentile increases bank flows to the recipient country by about 80 percent (distance-weighted regional neighbor measure).
- Neighbor inflow controls coefficients (Table 7 highlights) are positive and significant in many columns, e.g., 3.840** (col. (1)), 5.790*** (col. (8)).
- Including neighbor inflow controls does not materially change estimated effect of recipient country CARs.
- Implication: potential for "capital control wars" and need for multilateral coordination among recipient countries.

### Sensitivity analysis and endogeneity checks
- Robustness to inclusion of additional controls: institutional quality, financial development (stock market capitalization, private credit, deposit money bank assets to GDP), financial soundness (bank ROA, ROE), bank concentration, bank z-score, common lender contagion, bilateral trade, and direct exchange rate relationships — CAR coefficients remain similar.
- Alternate dependent variables:
  - Using (log) real flows (deflated by US CPI) or bilateral stock of bank assets strengthens results in many cases.
  - Moving from 25th to 75th percentile on outflow measures reduces bilateral stock by about 12-15 percent; similar movement on inflow measures reduces bilateral stock by about 8-23 percent.
- Nonbank sector flows: financial-sector specific measures less related to flows to nonbanking sector; bond inflow controls associated with lower inflows to nonbanks.
- Alternate CAR definitions: excluding mild formalities yields similar results; direct investment inflow controls become significant at 10 percent with alternative coding.
- IV-2SLS (instruments: monetary freedom; left-wing government):
  - First-stage: monetary freedom negatively and significantly associated with CARs; F-tests and R-squared (~0.9) indicate instrument strength.
  - Second-stage: predicted outflow and inflow controls imply reductions in flows of about 50-100 percent (outflows) and 50-85 percent (inflows) moving from lower to top quartile.
  - Example second-stage coefficients (Panel B notes): Overall controls (IV) -3.222***; Bond controls -1.570*; Direct investment controls -4.347***; Financial credit controls -1.567***; Lending to nonresidents -6.243***; Lending locally in FX -1.922***.
  - Overidentification / excludability check: monetary freedom not significant in second stage when left-wing government used as instrument.

### Policy implications and recommendations
- CARs in both source and recipient countries can significantly influence cross-border bank flows.
  - Source-side effective measures: controls on bond, equity, direct investment and financial credit outflows; restrictions on lending to nonresidents.
  - Recipient-side effective measures: bond inflow controls; FX-related prudential measures (restrictions on local FX lending; open FX position limits).
- Joint use of measures at both ends can:
  - Produce larger reductions in flows; or
  - Achieve the same reduction with less intensive measures at each end (potential global efficiency if costs are convex).
- Where administrative capacity or treaty obligations limit action at one end, relevant CARs at the other end can still reduce flows.
- Given the role of source-country monetary policies in pushing bank flows to recipient countries, coordination on monetary policy may enhance effectiveness of domestic countercyclical measures and reduce adverse cross-border spillovers.
- Presence of spillovers strengthens the case for coordination among recipient countries to avoid beggar-thy-neighbor outcomes or costly "capital control wars."
- Further research recommended to:
  - Better measure intensity of CARs across countries;
  - Analyze interactions across different types of measures and other types of capital flows;
  - Delineate spillovers across measure types and operationalize international cooperation.

*Source: _wp14188 - Section 4 analyzes the implications of implementing CARs simultaneously in both source (IMF working paper section content provided).*

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

### _wp14188 - References .............................................................................................................

### Introduction and research questions
- Investigates whether adoption of capital controls or prudential measures (collectively, capital account restrictions, CARs) on outflows by source countries and on inflows by recipient countries can moderate large and potentially disruptive cross-border bank flows.
- Motivated by debates following the global financial crisis (GFC) on whether to regulate capital flows at both source and recipient ends and whether coordinated action could improve outcomes.
- Uses bilateral cross-border bank flow data to assess the joint effectiveness and interaction of CARs at both ends.

### Data, sample, and measurement
- Bilateral data on cross-border bank flows from 31 major source to 76 major recipient countries over 1995–2012.
- CARs constructed using detailed information from the IMF’s Annual Report on Exchange Arrangements and Exchange Restrictions (AREAER) and the OECD Code of Liberalization of Capital Movements.
- Analysis focuses on bank-intermediated credit flows and includes both residency-based capital controls and prudential measures (including foreign currency related prudential measures).

### Key empirical findings
- Association between CARs and lower volumes of cross-border bank flows:
  - On the source side: moving from the lower to the upper quartile on measures of overall, bond, equity, direct investment or financial credit outflow controls is associated with about 50-100 percent lower flows.
  - Prudential measures such as restrictions on lending to nonresidents imply a similar reduction in flows.
  - On the recipient side: moving from the lower to the upper quartile on overall and bond inflow controls, and on the existence of foreign currency (FX) related prudential measures (such as restrictions on lending in FX, and open FX position limits) is associated with some 50-80 percent lower inflows.
- Global and domestic determinants:
  - Strong effect of global risk aversion and source-country monetary policy (proxied by the interest rate) on bank flows, highlighting procyclicality.
  - Recipient-country domestic interest rate and exchange rate regime also materially affect flows.
- Joint operation of source and recipient CARs:
  - Estimated effects remain largely unchanged when controlling simultaneously for both sides’ CARs.
  - Individual measures are associated with a larger reduction in flows when the other side is financially more open (though not necessarily fully open).
  - Source and recipient restrictions are partially (but not fully) additive: coordination can achieve either a larger reduction in flows or the same reduction with less intensive measures at each end.

### Regional and spillover results
- Pre-GFC flows to emerging Europe and the Eurozone peripheral countries would have been substantially lower in the presence of CARs (including prudential measures) at either end.
- Evidence of potential spillover effects from inflow controls:
  - The volume of bank flows received by a recipient country is significantly larger when its neighbors (defined by regional or economic similarity) are financially relatively closed—indicating cross-border spillovers and potential for “capital control wars.”

### Robustness and identification
- Results survive robustness tests, including:
  - Addressing endogeneity via instrumental variable (IV) approach using (lack of) monetary/central bank freedom and the presence of a democratic left-wing government as instruments for the existence of CARs.
  - Using alternate samples: restricting source (recipient) sample to advanced (EM) countries; excluding post-GFC years; removing offshore financial centers.
  - Defining dependent variable in alternate ways (e.g., in stock rather than flow terms).

### Policy implications and recommendations
- Adoption of relevant CARs in boom times could help dampen the procyclicality of cross-border bank flows, lowering the risk of systemic financial crises.
- Coordination between source and recipient countries could be useful to achieve a given reduction in cross-border flows by adopting relatively lower levels of restrictions at both ends—potentially a globally more efficient outcome when the cost function of capital controls is convex.
- Coordination may be especially beneficial when capacity or legal constraints limit action at one end.
- Given the role of source-country monetary policies in pushing bank flows to recipient countries, coordination on monetary policy may enhance effectiveness of domestic countercyclical measures and reduce adverse cross-border spillovers.
- Considering spillover effects of inflow-related CARs, coordination among recipient countries could avoid costly “capital control wars.”

### Contributions relative to existing literature
- Examines the joint effect of outflow restrictions by source countries and inflow restrictions by recipient countries, rather than focusing on one side only.
- Analyzes the impact of CARs on banking flows—an increasingly important and volatile component of total cross-border capital flows—rather than only aggregate or portfolio flows.
- Constructs a comprehensive bilateral dataset of capital controls (disaggregated by asset type) and prudential measures for both source and recipient countries over 1995–2012, enabling identification of significant associations and spillover effects.

_Italic: Source: _wp14188 - References ............................................................................................................._

### Section 4 analyzes the implications of implementing CARs simultaneously in both source

### Section 4 analyzes the implications of implementing CARs simultaneously in both source

### Data and measurement
- Annual bilateral bank asset flows from 31 reporting source countries to 76 recipient countries over 1995–2012 from BIS Locational Banking Statistics by Residence.
- Flows defined as exchange rate-adjusted changes in gross international financial claims of resident banks on bank and nonbank sectors of recipient countries; data cover over 90 percent of international assets of domestic banking institutions.
- CARs drawn from IMF AREAER and OECD Code of Liberalization of Capital Movements.
- Prudential measures captured as binary variables (1 = restriction present, 0 = otherwise).
- Capital control measures constructed following Schindler (2009); averages of binary indicators by transaction type, taking values 0, 0.5, or 1 for bond/equity controls and 0 or 1 for direct investment and financial credit where disaggregation is limited.
- Overall restrictiveness indices constructed as averages over asset classes and range between 0 and 1.
- Lagging of country-specific variables by one period to mitigate reverse causality; cluster standard errors at country-pair level.

### Empirical strategy
- Gravity-type panel regressions:
  - Equation (1): source-side outflow CARs and controls.
  - Equation (2): recipient-side inflow CARs and controls.
  - Equation (3): simultaneous inclusion of source outflow and recipient inflow CARs.
- Controls include (log) real GDP, real GDP per capita, real interest rates, real GDP growth, current account balance (percent of GDP), exchange rate regime (peg = 1), global VIX, commodity prices, and time and country-pair fixed effects.
- Robustness checks include OLS, CPFE/TE (country-pair fixed effects and time effects), alternative control sets, alternate dependent variables, alternate CAR definitions, IV-2SLS, and lag structures (1-year lag and 5-year lagged averages).

### Stylized facts (selected statistics)
- Total bank asset flows from reporting advanced source countries to advanced recipient countries:
  - about USD 435 billion in 1995;
  - almost USD 4 trillion in 2007;
  - about USD 280 billion in 2010 (post-GFC drop).
- On the eve of the crisis in 2007, bank inflows constituted:
  - about 50 percent of gross inflows to advanced countries;
  - about 40 percent of gross inflows to EMs.
- Proportion of EMs with restrictions on lending locally in FX:
  - about 50 percent in 2001;
  - 57 percent in 2007;
  - 70 percent in 2012.
- Proportion of EMs with restrictions on purchase of locally issued securities denominated in FX:
  - 42 percent in 2001;
  - 46 percent in 2007;
  - 49 percent in 2012.

### Main empirical findings — Source country (outflow) CARs
- Overall restrictiveness on capital outflows:
  - OLS: moving from the 25th to the 75th percentile associated with about 70 percent lower outflows.
  - CPFE/TE: same movement implies about 77 percent lower outflows.
- Disaggregated outflow controls (bond, equity, direct investment, financial credit):
  - Moving from the bottom to the top quartile reduces flows by some 50-100 percent.
- Prudential measure: restrictions on lending to nonresidents:
  - Associated with about 100 percent reduction in bank asset flows (statistically significant).
- Source country real interest rate:
  - A 100 basis points increase reduces bank outflows by about 20 percent (CPFE/TE).

### Main empirical findings — Recipient country (inflow) CARs
- Overall capital inflow controls (CPFE/TE):
  - Moving from the 25th to the 75th percentile associated with about 50 percent lower bank asset flows (significant at 5 percent).
- Bond inflow controls:
  - Moving from the 25th to the 75th percentile lowers bank inflows by about 64 percent.
- Equity and direct investment inflow controls:
  - Statistically insignificant in baseline CPFE/TE results.
- Financial-credit inflow controls:
  - Marginally insignificant (p-value = 0.11).
- FX-related prudential measures in recipients:
  - Restrictions on lending locally in FX and open FX position limits: inflows about 70-80 percent lower when present.

### Joint implementation and interaction of CARs
- Simultaneous inclusion of source outflow and recipient inflow CARs yields coefficients similar in magnitude and significance to individual estimates.
- Combined effects are not fully additive due to log specification; example:
  - Source overall outflow index: 25th to 75th percentile → about 80 percent reduction (holding recipient constant).
  - Recipient bond inflow index: 25th to 75th percentile → about 50 percent reduction (holding source constant).
  - Joint impact of both measures → 63 percent reduction.
- Effectiveness conditional on openness of the other end:
  - Outflow measures have larger impact when recipient country is more open (below 75th percentile on its inflow index).
  - Inflow measures have larger impact when source country is more open (below 75th percentile on its outflow index).
  - When the other side is already relatively closed, incremental impact is smaller and often statistically insignificant.
- Excluding fully open counterparties (no restrictions) does not materially change results.

### Counterfactual scenarios (using estimates from Table 4 col. [6])
- Simulated change in predicted pre-GFC (2007) flows under three scenarios using financial credit outflow controls (source) and restrictions on lending locally in FX (recipient):
  1. All source countries impose financial credit outflow controls; other variables at actual 2007 levels.
  2. All recipient countries impose restrictions on lending locally in FX; other variables at actual 2007 levels.
  3. Both all source and all recipient countries impose the respective measures; other variables at actual 2007 levels.
- Selected regional results for 2007:
  - Emerging Europe:
    - Scenario 1: about 20 percent lower flows.
    - Scenario 2: about 80 percent lower flows.
    - Scenario 3: about 85 percent lower flows.
  - Eurozone peripheral countries (Greece, Ireland, Italy, Portugal and Spain):
    - Scenario 3: flows might have been almost 95 percent lower.
  - Latin America:
    - Scenario 1: smaller effect because some source countries already had controls.
    - Scenario 2: flows lowered by about 80 percent.
  - Asia:
    - Scenario 1: source country action reduces flows by over 30 percent.
    - Scenario 2: smaller effect since many recipient Asian countries already had FX lending restrictions in 2007.
  - Post-GFC surge control:
    - Action by all source and recipient countries would have lowered inflows to Asian and Latin American EMs by about 10 and 90 percent, respectively.

### Spillover effects
- Evidence of capital flow deflection: raising average inflow controls in regional or economically similar neighbors increases bank flows to a recipient country.
  - Raising average inflow controls in regional neighbors from the median to the 75th percentile increases bank flows to the recipient country by about 80 percent (distance-weighted regional neighbor measure).
- Similar results when grouping by economic characteristics (k-means clustering) and using distance-weighted or GDP-weighted neighbor measures.
- Including neighbor inflow controls does not materially change estimated effect of recipient country CARs.
- Implication: potential for "capital control wars" and need for multilateral coordination among recipient countries.

### Sensitivity analysis and endogeneity checks
- Robustness to inclusion of additional controls:
  - Institutional quality, financial development (stock market capitalization, private credit, deposit money bank assets to GDP), financial soundness (bank ROA, ROE), bank concentration, bank z-score, common lender contagion, bilateral trade, and direct exchange rate relationships — CAR coefficients remain similar.
- Alternative dependent variables:
  - Using (log) real flows (deflated by US CPI) or bilateral stock of bank assets strengthens results in many cases; moving from 25th to 75th percentile on outflow measures reduces bilateral stock by about 12-15 percent; similar movement on inflow measures reduces bilateral stock by about 8-23 percent.
- Nonbank sector flows:
  - Financial sector specific measures less related to flows to nonbanking sector; bond inflow controls associated with lower inflows to nonbanks.
- Alternate CAR definitions (exclude mild formalities as restrictions) yield similar results; direct investment inflow controls become significant at 10 percent with alternative coding.
- Endogeneity:
  - Use of first lag and 5-year lagged average; IV-2SLS with instruments:
    - Monetary freedom (Heritage Foundation Economic Freedom Index) and presence of a left-wing government (Beck et al., 2001) as instruments.
    - First stage: monetary freedom negatively and significantly associated with CARs; F-tests and R-squared (~0.9) indicate instrument strength.
    - Second stage: results consistent with benchmark—predicted outflow and inflow controls imply reductions in flows of about 50-100 percent (outflows) and 50-85 percent (inflows) moving from lower to top quartile.
    - Overidentification / excludability check: monetary freedom not significant in second stage when left-wing government used as instrument, supporting validity.

### Policy implications and conclusions
- CARs in both source and recipient countries can significantly influence cross-border bank flows.
  - Source-side effective measures: controls on bond, equity, direct investment and financial credit outflows; restrictions on lending to nonresidents.
  - Recipient-side effective measures: bond inflow controls; FX-related prudential measures (restrictions on local FX lending; open FX position limits).
- Joint use of measures at both ends can:
  - Produce larger reductions in flows; or
  - Achieve the same reduction with less intensive measures at each end (potential global efficiency if costs are convex).
- Where administrative capacity or treaty obligations limit action at one end, relevant CARs at the other end can still reduce flows.
- Presence of spillovers strengthens the case for coordination among recipient countries to avoid beggar-thy-neighbor outcomes or costly "capital control wars."
- Further research recommended to:
  - Better measure intensity of CARs across countries;
  - Analyze interactions across different types of measures and other types of capital flows;
  - Delineate spillovers across measure types and operationalize international cooperation.

*Source: _wp14188 - Section 4 analyzes the implications of implementing CARs simultaneously in both source (IMF working paper section content provided).*

### REFERENCES

### _wp14188 - REFERENCES

### Bibliography and literature cited
- Comprehensive list of cited works spanning empirical and theoretical studies on capital flows, cross-border banking, capital controls, prudential measures, and financial globalization, including but not limited to:
  - Acemoglu, D., S. Johnson, and J. Robinson, 2001, “The Colonial Origins of Comparative Development: An Empirical Investigation,” American Economic Review, 91(5): 1369-1401.
  - Brunnermeier, M., et al., 2012, Banks and Cross-Border Capital Flows: Policy Challenges and Regulatory Responses (Washington DC: Brookings Institution).
  - Dell’Ariccia, G., et al., 2008, “Reaping the Benefits of Financial Globalization,” IMF Occasional Paper 264.
  - Ghosh, A., M. Qureshi, J. Kim, and J. Zalduendo, 2014a, “Surges,” Journal of International Economics, 92(2): 266-285.
  - Ostry, J., A. Ghosh, K. Habermeier, M. Chamon, M. Qureshi, D. Reinhardt, 2010, “Capital Inflows: The Role of Controls,” IMF Staff Position Note SPN/10/04.
- Note: Full reference list reproduced in source PDF.

### Figures and empirical illustrations (1995–2012)
- Figure 1: Cross-Border Bank Asset Flows, 1995–2012
  - Panels by source country type (advanced; EM) in USD bln. and percent of GDP.
  - Source: BIS Locational Statistics.
- Figure 2: Cross-Border Bank and Nonbank Liability Flows, 1995–2012 (in USD tril.)
  - Panels: Advanced countries; Emerging markets; Emerging Asia; Latin America; Emerging Europe; Baltics.
  - Source: BIS Locational Statistics and IFS database (BPM5 presentation).
- Figure 3: Capital Controls and Prudential Measures, 1995–2012 (in percent of observations)
  - Panels: (a) Inflow-related measures (recipient countries); (b) Outflow-related measures (source countries).
  - Source: IMF’s AREAER and the OECD Code of Liberalization of Capital Movements.
- Figure 4: Cross-Border Bank Flows, Capital Controls, and Prudential Measures, 1995–2012
  - Panels: (a) Capital controls; (b) Outflow prudential measures; (c) Inflow prudential measures.
  - Note: Banking flows measured as log of exchange rate adjusted changes in the total stock (amounts outstanding) of assets (all instruments). *** indicates statistically significant different means between the two groups at the 1 percent level.
- Figure 5: Potential Impact of CARs on Cross-Border Bank Flows (in percent)
  - Panels: (a) Pre-GFC capital flows, 2007; (b) Post-GFC capital flows, 2010.
  - Note: Uses estimates reported in Table 4 (col. [6]) and sums predicted values for all countries in the identified region. Shows changes in predicted flows (in percent) under scenarios where source and/or recipient countries impose specified measures.

### Key empirical findings and coefficients (selected, 1995–2012)
- Sample and estimation context (benchmark regressions in Tables 2–4):
  - Observations: 22,257.
  - No. of source (recipient) countries: 31 (76).
  - No. of country-pairs: 1,943.
  - Estimation: country-pair fixed effects / year effects included in many specifications.
  - R2 (within CPFE): 0.05 reported in many CPFE/TE specifications.
- Table 2 (Cross-Border Bank Flows and Source Country CARs):
  - Capital outflow controls (i): coefficient reported -2.442*** (col. (1)) and -2.978*** (col. (2)).  
  - Bond outflow controls (i): -1.446* (col. (3)).  
  - Equity outflow controls (i): -2.316*** (col. (4)).  
  - Direct investment outflow controls (i): -4.019*** (col. (5)).  
  - Financial credit outflow controls (i): -1.438*** (col. (6)).  
  - Lending to nonresidents (i): -5.505*** (col. (7)).  
  - Log (VIX): -7.933*** (reported).
  - Log (Distanceij): -1.374***.
  - Off-shore countriesij: 1.742***.
- Table 3 (Cross-Border Bank Flows and Recipient Country CARs):
  - Bond inflow controls (j): -2.023*** (col. (3)).  
  - Lending locally in FX (j): -1.792*** (col. (7)).  
  - Open FX position limits (j): -1.110* (col. (9)).  
  - Log (VIX): -7.716***.
- Table 4 (Source and Recipient CARs jointly):
  - Capital outflow controls index (i): -2.999*** (col. (1)); similar negative magnitudes across columns.  
  - Capital inflow controls index (j): -2.027** (col. (2)).  
  - Direct investment outflow controls index (i): -4.020***.  
  - Financial credit outflow controls index (i): -1.443***.  
  - Lending to nonresidents (i): -5.489***.  
  - Lending locally in FX (j): -1.801***.  
  - Open FX position limits (j): -1.116*.
- Table 5 & Table 6 (By openness):
  - Effects of outflow/inflow controls remain negative and often statistically significant when splitting samples into more open vs less open source/recipient countries; sample sizes vary (e.g., Observations 16,934 vs 5,323; 19,374 vs 2,883).
- Table 7 (Spillover Effects of CARs):
  - Inflow controls in neighbors j: positive and statistically significant coefficients in many columns, e.g., 3.840** (col. (1)), 5.790*** (col. (8)), indicating measurable spillovers from neighboring countries’ inflow controls.
  - Capital outflow controls index i: -2.992*** (col. (1)).  
  - Bond inflow controls index j: -1.857*** (col. (4)).  
  - Direct investment outflow controls i: -4.013*** (col. (5)).  
  - Lending to nonresidents i: -5.489*** (col. (6)).  
  - Lending locally in FX j: -1.720*** (col. (7)).
- Table 8 (Robustness analysis, benchmark vs alternatives):
  - Overall outflow restrictions (benchmark): -2.978*** (col. (1)).  
  - Lending to nonresidents remains strongly negative across robustness checks, e.g., -5.505*** (benchmark) and negative in alternative specifications.
  - Open FX position limit shows mixed signs across robustness checks (e.g., 0.419 in benchmark [not significant], negative and sometimes significant in some robustness columns).
  - Inflow restrictions (overall) show negative effects in many robustness columns, e.g., -1.998** (benchmark).
  - Country-pair/year fixed effects included across robustness specifications.
  - Observations vary across robustness columns (e.g., 22,257; 21,898; 20,656; 19,853; 18,989; alternative sample cuts such as Pre-GFC, Excl. offshore, etc.).
- Table 9 (IV-2SLS Estimates; first- and second-stage highlights):
  - First-stage: Monetary freedom is a statistically significant predictor for many CAR measures with coefficient -0.001*** or -0.002*** in Panel A (first-stage).
  - Second-stage (Panel B) reports IV-2SLS estimates where CA Rs are predicted by instruments; example second-stage coefficients: Overall controls (IV) -3.222*** (reported in Panel B notes), Bond controls -1.570*, Direct investment controls -4.347***, Financial credit controls -1.567***, Lending to nonresidents -6.243***, Lending locally in FX -1.922***.
  - Panel reports F-test (p-value) = 0.000 for instrument relevance; Observations ~22,257 (first-stage) and ~22,252 (second-stage variants).
- Appendix: Data and summary statistics (Table A2, Table A3)
  - Dependent variable: Bank asset flows from country i to j — exchange rate-adjusted change in external position of reporting banks by residence, expressed in log of USD.
  - Summary statistics (Table A3):
    - Bank asset flowsij (log): Obs 22,257; Mean 2.5215; Median 5.2018; Std dev 8.37; Min -26.442; Max 26.28.
    - Real GDPi (log): Obs 22,257; Mean 27.11; Median 27.14; Std dev 1.41; Min 23.28; Max 30.26.
    - Real GDPj (log): Obs 22,257; Mean 25.91; Median 25.89; Std dev 1.67; Min 21.54; Max 30.26.
    - Overall outflow controls index i: Obs 22,257; Mean 0.29; Median 0.25; Std dev 0.29; Min 0.00; Max 1.00.
    - Overall inflow controls index j: Obs 22,257; Mean 0.36; Median 0.25; Std dev 0.30; Min 0.00; Max 1.00.
    - Lending locally in FX j: Obs 22,257; Mean 0.36; Median 0.00; Std dev 0.48; Min 0.00; Max 1.00.
  - Data sources include BIS Locational Statistics, IMF WEO, IMF’s AREAER, OECD Code of Liberalization of Capital Movements, IFS, GDFR, Heritage Foundation, ICRG, Polity IV, DOTS, Bloomberg.
- Appendix: Sample composition (Table A1, Table A4)
  - Source countries (reporting): 31 listed (years in parentheses indicate reporting availability).
  - Recipient countries: 76 countries included (listed in Table A1).
  - Recipient country groupings by region and economic characteristics detailed in Table A4.

### Core empirical implications (as embodied in tables/figures)
- Source-country outflow restrictions (overall index and specific measures such as direct investment outflow controls, equity/bond/financial credit outflow controls, and measures restricting lending to nonresidents) are consistently associated with sizable and statistically significant reductions in bilateral cross-border bank asset flows (log of flows), with magnitudes often in the range of roughly -1.4 to -5.5 for specific measures and about -3.0 for aggregate outflow controls index in benchmark specifications (1995–2012).
- Recipient-country inflow restrictions (notably bond inflow controls, lending locally in FX, and open FX position limits) are associated with reductions in incoming cross-border bank flows; bond inflow controls show particularly robust negative coefficients (e.g., -2.023***).
- Spillover effects are present: inflow controls in neighboring or economically similar countries exhibit positive coefficients in some specifications (e.g., 3.840**), indicating cross-border reallocation or diversion effects; however, spillovers of source-country outflow controls remain negative domestically.
- Robustness checks (multiple specifications, sub-samples, IV-2SLS using instruments such as Monetary freedom and Left-wing government) broadly confirm the direction and significance of key relationships, with variations in magnitude across alternative samples and measures.
- Market volatility (Log VIX) is strongly negatively associated with bilateral bank flows (e.g., -7.933*** in reported specifications).

*Italicized source: Excerpted tables, figures, references, and appendix from the PDF chapter _wp14188 - REFERENCES (1995–2012 sample reporting as provided in source content).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2014/_wp14188.pdf_