Capital Flow Deflection Paolo Giordani (Luiss), Michele Ruta (IMF), - - PowerPoint PPT Presentation

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

Capital Flow Deflection

Paolo Giordani (Luiss), Michele Ruta (IMF), Hans Weisfeld(IMF), Ling Zhu(UMD) May 12, 2014

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

AVERAGE GROSS INFLOWS (% GDP) TO DEVELOPING COUNTRIES BY REGION

!5# 0# 5# 10# 15# 20# 25# 30# 35# 1995# 1997# 1999# 2001# 2003# 2005# 2007# 2009# La,n#America# Middle#East#and#North# Africa# Asia# Sub!Sahara#Africa# Former#Soviet#Bloc# Central#and#East#Europe#

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

MOTIVATION

◮ Capital inflows to developing countries are increasingly

large and volatile.

◮ Capital controls can be used as prudential policy ◮ Open questions:

◮ what are the spillover effects of capital controls? ◮ what multilateral institutions do countries need to govern

capital flows?

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

BRAZIL’S CONTROLS AND SOUTH AFRICA’S FLOWS

0" 0.2" 0.4" 0.6" 0.8" 1" 1.2" )2" 0" 2" 4" 6" 8" 10" 12" 14" 1990" 1991" 1992" 1993" 1994" 1995" 1996" 1997" 1998" 1999" 2000" 2001" 2002" 2003" 2004" 2005" 2006" 2007" 2008" 2009" 2010" 2011" 2012" South"Africa's"Gross"Inflows"(%"GDP)" South"Africa's"Inflow"Controls" Brazil's"Inflow"Controls"

Note: The left vertical axis provides scale for gross capital inflow as percentage of GDP. The vertical right axis provides the scale for Schindler’ measure of inflow controls.

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

THIS PAPER

◮ Start from a simple two-period model of optimal capital

controls

◮ Capital flow deflection: inflow restrictions in some

countries can increase capital inflows to others

◮ Policy response: controls can trigger more controls

◮ Investigate empirical evidence of these multilateral effects

◮ Using panel data for a large sample of LIC/EM, 1995-2009

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

CONTRIBUTION

◮ We formalize capital flow deflection and the subsequent

policy response.

◮ We provide the first cross sectional empirical evidence for

spillover effects of capital controls.

◮ We find strong evidence of capital flow deflection to

countries with similar economic characteristics.

◮ Notwithstanding these strong cross-border spillover

effects, we find no evidence of policy response.

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

RELATED LITERATURE

◮ Theory of optimal capital controls: Korinek (2014),

Costinot et al. (2013)

◮ Empirical evidence of spillover effect of capital controls:

Forbes et al. (2012), Lambert et al. (2012)

◮ Multilateral rules for capital controls: Jeanne (2012), Ostry

et al. (2012)

◮ Determinants of capital flows: Forbes and Warnock (2012),

Ghosh et al. (2013)

◮ Determinants of capital controls: Fratzscher (2012),

Fernandez et al. (2013)

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

ROADMAP

◮ A simple model of optimal capital controls ◮ Empirical evidence of multilateral effects of capital controls ◮ Conclusion

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

Model

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

A SIMPLE TWO-PERIOD MODEL

◮ Multi-country world (i=1,.. . ,n) lasting for two periods

(t = 1, 2).

◮ Endowment of tradable goods yi t. ◮ Unit mass of identical consumers in country i maximize

Ui(ci

1, ci 2) = u(ci 1) + βiu(ci 2) + e(Ci 1 − Yi 1),

s.t. (ci

1 − yi 1)(1 + τ i)R = yi 2 − ci 2 + Ti

where τ i is the capital control in country i.

◮ e(·) captures the negative externality associated with

aggregate capital inflows, e.g. financial fragility

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

SOCIAL PLANNER

◮ Social planner in country i chooses {ci 1, ci 2} to maximize the

social welfare function Wi

t(ci 1, ci 2) = u(ci 1) + βiu(ci 2) + e(ci 1 − yi 1), ◮ Individual consumers do not internalize the negative

externality while the social planner does.

◮ The externality provides an efficiency rationale for capital

controls.

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

OPTIMAL POLICY

◮ Unilaterally optimal capital controls are:

1 + τ i∗ = (1 + ˆ τ i)(1 + ˜ τ i)

◮ Prudential motive:

• τ i =
• x′(ci

1−yi 1)

u′(ci

1)−x′(ci 1−yi 1)

when ci

1 − yi 1 > 0

when ci

1 − yi 1 ≤ 0.

◮ Terms of trade motive:

˜ τ i = miǫ−i, where mi is the size of country i. ǫ−i is the inverse elasticity of global savings faced by i.

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

DEFINITION OF WORLD MARKET EQUILIBRIUM

◮ A world market equilibrium is defined as the gross real

world interest rate R∗ and each country’s consumption plan and capital controls {ci

1, ci 2, τ i∗}n i=1 that satisfy

u′(ci

1) = βiR∗u′(ci 2)(1 + τ i∗),

(ci

1 − yi 1) = (yi 2 − ci 2)

R∗ ,

n

• i=1

mi(yi

1 − ci 1) = 0,

1 + τ i∗ = (1 + τ i)(1 + τ i).

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

CAPITAL FLOW DEFLECTION

◮ Let Ω denote the set of borrowing countries, and group

S ⊆ Ω with i belonging to group S.

◮ A rise in the capital controls in the group S−i = S \ {i}

causes an increase in capital inflows to country i: d

• ci

1(R, τ i) − yi 1]

dτ S−i > 0

◮ Intuition: capital controls create a spillover effect on other

borrowers

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

CAPITAL FLOW DEFLECTION

R" R" R" World"market" Borrowing"countries"S6i" Borrowing"country"i" R*" R*’" A" B" C" D" E" F" B’" D’" F’" τS6i"

CFS" CFD" CFD’" CFDi" CFDS6i"

DD’ is the capital flows deflected to country i.

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

POLICY RESPONSE

◮ A rise in capital controls in the set of countries S−i induces

policy response in country i such that: dτ i∗ dτ S−i = d τ i dτ S−i (1 + τ i) + d τ i dτ S−i (1 + τ i), where (i) d τ i dτ S−i = d τ i dR dR∗ dτ S−i > 0 and (ii) d τ i dτ S−i = − mi dε−i

dτ S−i

mi dε−i

d τ i − 1

≶ 0.

◮ Intuition: capital flow deflection induces more prudential

capital controls.

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

MULTIPLIER EFFECT

◮ A shock to the set of countries S ⊆ Ω causes a prudential

policy response of each country i ∈ S such that |d τ i dρ | = |∂ τ i ∂ρ + d τ i dR dR∗ dτ S · d τ S dρ | > |∂ τ i ∂ρ |

◮ Intuition: the complementarity between

τ and R∗ creates a chain reaction amplifying the policy response to the shock.

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

MULTIPLIER EFFECT

R

E

b

b›R,_fi b›R,_vfi RD›bfi

E’’ E’ F E’’’

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

Do Capital Controls Deflect Capital Flows and Lead to a Policy Response?

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

DATA

◮ Sample countries:

◮ Focus on LIC/EM where capital controls are used as a

policy tool

◮ 78 countries−select those with at least 10 years of

• bservations for all variables

◮ Country sample is larger (131) when computing τ S−i.

◮ Sample period:

◮ Annual data from 1995-2009 ◮ Constrained by the inflow control data (Schindler, 2009)

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

EMPIRICAL STRATEGY: CAPITAL FLOW DEFLECTION

◮ Pull-push model with spillover variable:

ωi

t = β0 + β1Rt + β2τ i t−1 + β3τ S−i t

+ β4xi

t + ui t

ωi

t denotes country i’s gross inflows as a share of GDP.

τ S−i

t

denotes the spillover variable–inflow controls in the set of countries S−i; it is computed as τ S−i

t

=

• j∈S−i yj

tτ j t

• j∈S−i yj

t

yi

t is real GDP, xi t is a vector of the rest pull-push factors,

and ui

t is the error term.

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

CAPITAL FLOW DEFLECTION (S−i IS THE ENTIRE

SAMPLE EXCLUDING i)

(1) (2) Expected Signs Global Push Factors Real US interest rate

• 0.337**
• (0.152)

VIX

• 0.132***
• (0.0490)

Domestic Pull Factors (all lagged) Real GDP growth rate 0.595*** 0.504** + (0.188) (0.194) Real GDP growth rate shock

• 0.434**
• 0.406*
• (0.209)

(0.207) Real GDP per capita (logged) 0.652 0.481 + (0.571) (0.565) De jure Capital inflow control

• 2.468*
• 2.455*
• (1.341)

(1.440) Composite risk index 0.150** 0.143** + (0.0647) (0.0646) Spillovers ROG's inflow control 23.00*** 8.013 + (5.807) (27.50) Year fixed effects No Yes Observations 1,007 1,007 R-squared 0.135 0.167

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

NO CAPITAL FLOW DEFLECTION?

◮ No spillover effect is found when S−i is the entire sample

excluding i.

◮ Explanation: not all countries are close substitutes from

international investors’ perspective.

◮ Solution: divide countries into groups of likely substitutes

based on common characteristics:

◮ Geographic location (WEO), 6 groups. ◮ Export specialization (WEO), 5 groups. ◮ Return: growth rate (WEO), 4 groups. ◮ Risk: composite risk index (PRS), 4 groups.

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

WITHIN-GROUP CAPITAL FLOW DEFLECTION

(1) (2) (3) (4) (5) (6) Group by geographic location

• 4.476

1.135

• 0.614
• 3.801

1.723

• 1.935

(2.793) (3.922) (4.137) (2.872) (4.582) (4.324) Group by export specialization

• 1.057

5.377 4.090

• 0.916

5.192 3.770 (2.614) (4.111) (4.208) (2.719) (3.952) (4.089) Groups by returns Time-invariant groups by growth rate 2.873 11.43* 11.60* 1.920 6.248 5.967 (3.743) (5.934) (6.583) (3.477) (4.342) (4.380) Time-variant groups by growth rate 1.141

• 2.261
• 2.643
• 0.114
• 2.263
• 1.926

(2.163) (1.812) (1.634) (2.067) (1.730) (1.345) Groups by risks Time-invariant groups by composite risk 3.390 10.71** 19.26*** 0.0809 1.380 6.818 (4.599) (4.992) (5.056) (4.190) (4.477) (4.265) Time-variant groups by composite risk 7.719*** 3.416* 3.356* 8.480*** 5.535** 4.531** (2.830) (2.032) (1.744) (3.087) (2.267) (1.877) Year FE Yes Yes Yes Yes Yes Yes Region FE No Yes Yes No Yes Yes Group FE No Yes Yes No Yes Yes Contagion variables No Yes Yes No Yes Yes Country FE No No Yes No No Yes Lagged ROG's inflow control No No No Yes Yes Yes

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

DISCUSSION

◮ In the previous table, we only report coefficients for the

spillover variable τ S−i.

◮ There are 36 specifications which differ by country groups

(6 rows) and by the set of controls (6 columns).

◮ We find strong evidence of within-group capital flow

deflection when countries are grouped based on risk:

◮ In 2009 restrictions in Brazil are estimated to have increased

flows to South Africa by 0.42 − 0.95% of GDP.

◮ We find no capital flow deflection on neighbors.

◮ Possible explanation: investors expect policy contagion.

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

SPILLOVERS FROM CAPITAL ACCOUNT LIBERALIZATION IN CHINA AND INDIA

◮ Removal of inflow restrictions in China and India would

reduce gross private inflows to countries in the same risk group.

!! !! Reduction of Private Gross Inflow as Percentage of GDP Impact of China's Capital Account Liberalization Chile Mexico Malaysia Russian Poland Lower Bound Estimate

• 1.72
• 1.85
• 1.72
• 1.92
• 1.76

Upper Bound Estimate

• 4.33
• 4.67
• 4.34
• 4.85
• 4.45

Impact of India's Capital Account Liberalization South Africa Argentina Brazil Indonesia Thailand Lower Bound Estimate

• 0.72
• 0.72
• 0.95
• 0.75
• 0.71

Upper Bound Estimate

• 1.81
• 1.82
• 2.40
• 1.90
• 1.80

Note: The counterfactual exercise considers the total elimination of inflow restrictions in India and China in 2009.

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

ROBUSTNESS TEST

◮ Gross inflow surges: Ghosh et al. (2012). ◮ Alternative measures of capital controls: Quinn et al.

(2011), Ostry et al. (2012).

◮ Alternative measures of country risks: law and order,

property rights.

◮ IV approach: instrumenting τ S−i with upcoming election

dummy and average regional restrictions.

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

ALTERNATIVE MEASURES OF CAPITAL CONTROLS

(1) (2) (3) (4) (5) (6) CAPITAL 19.94*** 14.20** 18.38*** 11.56** 8.161 8.218 (5.840) (5.397) (6.350) (5.192) (5.119) (7.100) FINCONT2 6.875** 7.155** 4.059 8.702** 10.54*** 7.204** (3.226) (2.645) (2.580) (3.453) (3.350) (2.813) Year FE Yes Yes Yes Yes Yes Yes Region FE No Yes Yes No Yes Yes Group FE No Yes Yes No Yes Yes Contagion variables No Yes Yes No Yes Yes Country FE No No Yes No No Yes Lagged ROG's inflow control No No No Yes Yes Yes

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

IV RESULT

(1) (2) Second Stage Rest of the Group's Inflow Control 10.746*** 9.467* (4.035) (5.604) First Stage Upcoming Election Instrument 0.0886*** 0.161*** (0.0205) (0.0370) Trade Agreement Instrument

• 0.298***

(0.0224) Right-Wing Government Instrument

• 0.242***

(0.0392) First Stage F statistics for the instruments 74.590 16.002 P-value of F statistics 0.000 0.000 Hansen J statistics 0.201 0.449 P-value of Hansen J statistics 0.654 0.503 Year FE Yes Yes Group Contagion Yes Yes Country FE Yes Yes Observations 879 450 Countries 68 35

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

EMPIRICAL STRATEGY: POLICY RESPONSE

◮ Probit model of increase in inflow controls:

Pr(Inci

t = 1) = Φ(β0 + β1∆τ S−i t

+ β2xi

t + ui t)

Inci

t equals 1 if country i raises inflow controls at time t and

0 otherwise. ∆τ S−i

t

is the change in the spillover variable τ S−i

t

. xi

t includes all the other pull and push factors, fixed effects,

regional/group controls. xi

t also includes additional determinants of capital controls

discussed in the literature.

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

WITHIN-GROUP POLICY RESPONSE

(1) (2) (3) (4) (5) (6) Expected Sign Domestic Variables (Lagged) Real GDP growth rate

• 0.0533**
• 0.0593**
• 0.0339
• 0.0815**
• 0.0628**
• 0.0513*

+ (0.0267) (0.0273) (0.0320) (0.0329) (0.0301) (0.0266) Real GDP growth rate shock 0.0825** 0.0875** 0.0587 0.114*** 0.0885** 0.0804**

• (0.0352)

(0.0351) (0.0383) (0.0440) (0.0367) (0.0347) REER overvaluation

• 1.197**
• 1.148**
• 1.090*
• 1.238**
• 1.343**
• 1.186**
• (0.544)

(0.544) (0.568) (0.534) (0.541) (0.559) Inflation 0.229** 0.220* 0.224** 0.220** 0.250** 0.233** + (0.112) (0.113) (0.113) (0.110) (0.109) (0.112) Flexible Exchange Rate

• 0.0855
• 0.0824
• 0.116
• 0.0539

0.0159

• 0.0462
• (0.209)

(0.217) (0.164) (0.180) (0.207) (0.223) Real GDP (logged) 0.0840* 0.0914* 0.108** 0.0745 0.0828* 0.0837* + (0.0455) (0.0480) (0.0485) (0.0455) (0.0461) (0.0458) External Debt

• 0.206
• 0.138
• 0.214
• 0.225*
• 0.190
• 0.191

+ (0.129) (0.131) (0.131) (0.123) (0.143) (0.128) Change of Weighted Inflow Controls in the ROG Group by geographic location

• 1.065

+ (1.119) Group by export specialization 0.646 + (0.975) Time-invariant group by growth rate

• 0.142

+ (1.148) Time-variant group by growth rate

• 0.317

+ (0.295) Time-invariant group by composite risk

• 0.901

+ (1.315) Time-variant group by composite risk

• 0.0150

+ (0.380) Observations 836 836 836 836 836 836 Pseudo R-squared 0.151 0.151 0.150 0.150 0.150 0.150

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

DISCUSSION

◮ We find no evidence of a policy response even within risk

groups despite the large capital flow deflection.

◮ These negative findings are resilient to robustness tests:

◮ Drop large economies to remove possible TOT motive ◮ Linear probability model ◮ Alternative definition of the spillover variable ◮ Decrease in capital controls as dependent variable

◮ This lack of evidence may be driven by:

◮ Data methodology: capital control measures capture only

the extensive margin of restrictions.

◮ Historical period: until recently capital controls were not

considered “legitimate” prudential policies (IMF, 2012).

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

CONCLUSION

◮ The paper presents a simple model of capital flow

deflection and policy response.

◮ It provides evidence that inflow controls deflect capital to

countries with similar economic characteristics.

◮ We find no evidence of a policy response.

◮ However the change in view on prudential controls may

lead to policy response and multiplier effect in the future.

◮ Our findings inform the discussion on multilateral rules on

capital controls.

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

SUMMARY STATS

Variable Obs Mean S.D. Min Max Schindler index (inflows restrictions) 1170 0.49 0.34 0.00 1.00 Gross private inflows (% GDP) 1100 5.35 7.15

• 33.80

61.72 VIX 1170 21.54 6.45 12.58 34.04 Real US interest rate 1170 0.94 1.63

• 2.36

3.18 Inflation 1135 15.01 54.01

• 9.86

1061.21 Real GDP growth rate 1169 4.12 4.72

• 24.79

62.19 Nominal GDP per capita in USD (logged) 1159 7.42 1.17 4.38 10.66 Nominal GDP in USD (logged) 1170 10.10 1.70 6.47 15.42 Real effective exchange rate 1170 106.78 33.93 56.09 597.37 Composite risk 1138 66.24 8.72 28.00 86.70 Law and order 1138 3.38 1.14 1.00 6.00 Property rights 1121 0.44 0.17 0.05 0.90 de facto exchange rate regime 1134 2.17 1.15 1.00 6.00

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

DATA SOURCE

Data Source

Capital flows International Financial Statistics (IFS) Capital controls Schindler (2009) VIX Yahoo Finance US three month treasury rate Federal Reserve Economic Data (FRED) Inflation rate World Economic Outlook (WEO) Real GDP growth rate WEO Nominal GDP per capita in US$ WEO Nominal GDP in US$ WEO Real Effective Exchange Rate Information Notice System (INS) Composite risk index Political Risk Service (PRS) Law and Order PRS Property Rights Heritage Foundation de facto Exchange rate regime classification Iltzetzki et al. (2010)

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

TIME-INVARIANT GROUPS BY RISK

*'Composite'risk'index'is'used'to'measure'risk.'All'countries'listed'are'used'for'compu<ng'''''''''.'Countries'in'bold'are'used'for'regressions.'' High Risk High-Moderate Risk Low-Moderate Risk Low Risk Angola Albania Argentina Bahamas, The Burkina Faso Algeria Azerbaijan Bahrain Congo, Republic of Armenia Bolivia Botswana Côte d'Ivoire Bangladesh Brazil Chile

• Dem. Rep. of the Congo

Belarus Bulgaria China Ethiopia Cameroon Dominican Republic Costa Rica Guinea Colombia Egypt Croatia Guinea-Bissau Ecuador El Salvador Hungary Haiti Ghana Gabon Jordan Iraq Guyana Guatemala Kuwait Lebanon Honduras India Latvia Liberia Indonesia Iran Libya Malawi Kenya Jamaica Lithuania Mozambique Madagascar Kazakhstan Malaysia Myanmar Mali Papua New Guinea Mexico Nicaragua Moldova Paraguay Morocco Niger Mongolia Peru Namibia Nigeria Romania Philippines Oman Pakistan Senegal Russia Panama Sierra Leone Sri Lanka South Africa Poland Sudan Suriname Syria Qatar Togo Tanzania Thailand Saudi Arabia Turkey Uganda Ukraine Trinidad and Tobago Zambia Venezuela Uruguay Tunisia Zimbabwe Yemen Vietnam United Arab Emirates

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

GROSS INFLOWS BY COUNTRY GROUPS

• A. Country Groups by Regions
• B. Country Groups by Export Specialization
• C. Country Groups by Growth Rate
• C. Country Groups by Composite Risk Level

/5% 0% 5% 10% 15% 20% 25% 30% 35% 1995%1997%1999% 2001%2003%2005%2007%2009% La9n%America% Middle%East%and% North%Africa% Asia% Sub/Sahara%Africa% Former%Soviet%Bloc% /5% 0% 5% 10% 15% 20% Fuel%Exporter% Manufacturing% Goods%Exporter% Primary%Goods% Exporter% Service%Exporter% Diversified% Exporter% /5% 0% 5% 10% 15% 20% Slow%growth% Medium/slow% growth% Medium/fast% growth% Fast%growth% /5% 0% 5% 10% 15% 20% High%risk% Medium/high%risk% Medium/low%risk% Low%risk%

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

INFLOW CONTROLS BY COUNTRY GROUPS

0.2& 0.3& 0.4& 0.5& 0.6& 0.7& 0.8& 0.9& La=n&America& Middle&East&and& North&Africa& Asia& SubISahara&Africa& Former&Soviet& Bloc& 0.25& 0.3& 0.35& 0.4& 0.45& 0.5& 0.55& 0.6& 0.65& 0.7& 0.75& Fuel&Exporter& Manufacturing& Goods&Exporter& Primary&Goods& Exporter& Service&Exporter& Diversified& Exporter& 0.3& 0.35& 0.4& 0.45& 0.5& 0.55& 0.6& 0.65& Slow&growth& MediumIslow& growth& MediumIfast& growth& Fast&growth& 0.25& 0.3& 0.35& 0.4& 0.45& 0.5& 0.55& 0.6& 0.65& 0.7& High&risk& MediumIhigh&risk& MediumIlow&risk& Low&risk&

• A. Country Groups by Regions
• B. Country Groups by Export Specialization
• C. Country Groups by Growth Rate
• C. Country Groups by Composite Risk Level

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

WITHIN-GROUP CAPITAL FLOW DEFLECTION: SURGES

(1) (2) (3) (4) (5) (6) Group by geographic location

• 4.476

1.135

• 0.614
• 3.801

1.723

• 1.935

(2.793) (3.922) (4.137) (2.872) (4.582) (4.324) Group by export specialization

• 1.057

5.377 4.090

• 0.916

5.192 3.770 (2.614) (4.111) (4.208) (2.719) (3.952) (4.089) Groups by returns Time-invariant groups by growth rate 2.873 11.43* 11.60* 1.920 6.248 5.967 (3.743) (5.934) (6.583) (3.477) (4.342) (4.380) Time-variant groups by growth rate 1.141

• 2.261
• 2.643
• 0.114
• 2.263
• 1.926

(2.163) (1.812) (1.634) (2.067) (1.730) (1.345) Groups by risks Time-invariant groups by composite risk 3.390 10.71** 19.26*** 0.0809 1.380 6.818 (4.599) (4.992) (5.056) (4.190) (4.477) (4.265) Time-variant groups by composite risk 7.719*** 3.416* 3.356* 8.480*** 5.535** 4.531** (2.830) (2.032) (1.744) (3.087) (2.267) (1.877) Year FE Yes Yes Yes Yes Yes Yes Region FE No Yes Yes No Yes Yes Group FE No Yes Yes No Yes Yes Contagion variables No Yes Yes No Yes Yes Country FE No No Yes No No Yes Lagged ROG's inflow control No No No Yes Yes Yes

INTRODUCTION MODEL EMPIRICAL CONCLUSION APPENDIX

WITHIN-GROUP POLICY RESPONSE: LINEAR PROBABILITY MODEL WITH COUNTRY FE

(1) (2) (3) (4) (5) (6) Expected Sign Domestic Variables (Lagged) Real GDP growth rate

• 0.00264
• 0.00386
• 0.00290
• 0.00815
• 0.00335
• 0.00297

+ (0.00883) (0.00850) (0.00848) (0.00998) (0.00845) (0.00856) Real GDP growth rate shock 0.0110 0.0116 0.0109 0.0167 0.0105 0.0109

• (0.0105)

(0.0102) (0.0102) (0.0126) (0.0100) (0.00996) REER overvaluation

• 0.0340
• 0.0293
• 0.0344
• 0.0436
• 0.0447
• 0.0455
• (0.140)

(0.141) (0.141) (0.147) (0.141) (0.144) Inflation 0.0546 0.0501 0.0534 0.0530 0.0566 0.0548 + (0.0402) (0.0417) (0.0408) (0.0406) (0.0376) (0.0412) Flexible Exchange Rate

• 0.279***
• 0.288***
• 0.288***
• 0.291***
• 0.276***
• 0.321***
• (0.0465)

(0.0478) (0.0474) (0.0456) (0.0466) (0.0684) Real GDP (logged)

• 0.0432
• 0.104
• 0.0685
• 0.0715
• 0.109
• 0.0929

+ (0.382) (0.382) (0.380) (0.376) (0.371) (0.354) External Debt

• 0.0226
• 0.0175
• 0.0189
• 0.0224
• 0.0122
• 0.0193

+ (0.0369) (0.0376) (0.0369) (0.0382) (0.0384) (0.0382) Change of Weighted Inflow Controls in the ROG Group by geographic location

• 0.263

+ (0.234) Group by export specialization 0.0811 + (0.164) Time-invariant group by growth rate

• 0.0439

+ (0.240) Time-variant group by growth rate

• 0.0645

+ (0.0598) Time-invariant group by composite risk

• 0.138

+ (0.244) Time-variant group by composite risk 0.00406 + (0.0875) Observations 836 836 836 836 836 836 R-squared 0.166 0.167 0.165 0.168 0.169 0.168