are capital controls effective in preventing financial crisesa
TRANSCRIPT
Are Capital Controls Effective in Preventing Financial Crises?
Ling Huang
School of Economics, the Peking University and
Stanford Center for International Development
February 26, 2013
Fundamental questions: Capital controls Feb 18th 2010 The Economist
The IMF changes its mind on controls on capital inflows
Fundamental questions: Capital controls
• On April 5th 2011 the IMF released two documents designed to achieve just that … But several curious differences between them suggest that the fund’s own thinking on managing capital flows is far from settled.
• Jeanne, Subramanian and Williamson (2012) argue that capital control should not be a last resort, rather “…properly designed they might even be a regular instrument of economic policy.”
• Klein (2012):“ With a few exceptions, there is little evidence of the efficacy of capital controls on the growth of financial variables, the real exchange rate, or GDP growth at an annual frequency. These preliminary results raise doubts about assumptions behind recent calls for a greater use of episodic controls on capital inflows”
• Brazil, Iceland, Ireland, Peru and Turkey introduced new controls on capital inflows recently.
Key Points • data sample : 118 emerging market and developing
countries; time frame: 1970-2009 • Most currency crash events occurred in tight capital
control settings; • Loosening capital control is linked to reduced
likelihood of currency crises after controlling the relevant macroeconomic and financial variables.
• Preliminary evidence of discipline effect of capital account openness: in general, the tighter the capital control, the weaker the macroeconomic fundamentals.
Outline
• Key question: Upon controlling the usual macroeconomic and financial
correlates of currency crises, how does a higher degree of capital account openness marginally affect the likelihood of currency crises?
Outline • Structure :
– Literature review
– Identifying the currency crash events
– Univariate analysis: the behavior of the control variables in the antecedent (T-3: T) and aftermath (T+1:T+3) of the currency crash events – are they statistically different from the control group observations (“tranquil periods”)
– Multivariate analysis: Probit estimation
– Further empirical work to explain the Probit regression results.
Correlates of Currency Crises • We select the following correlates of currency crises suggested by the
literature : – (1) reserves adequacy variables: M2/international reserves, ratio
of short-term debt to international reserves – (2) Policy indicators: domestic credit growth, domestic credit to
private sector, M2 growth rate, budget balance (% of GDP), government consumption growth rate
– (3) Internal variables: inflation rate, real interest rate, investment growth, consumption growth, real growth rate
– (4) External variables: overvaluation, current account, export growth
– (5) External debt variables: total debt(% of GDP), short-term debt(% if total debt), short-term debt/export ratio, short-term debt (% of GDP), ratio of equity liabilities to debt liabilities
Index of de jure Financial Openness
The Chinn-Ito de jure index of financial openness kchin
variable min p25 p50 p75 max sd ------------------------------------------------------------------------------------- kchin -1.808 -1.129 -0.086 1.497 2.541 1.552
The Chinn-Ito Rating Index is constructed out of the natural break of the Chinn-Ito de jure index of capital control:
kchinr = 1, 2, 3, 4, 5
1 2 3 4 5 close open
Distribution of Countries under Different Degree of Financial Openness
• 118 emerging and developing countries, Year 2007
Number of countries under each level of kchinr
kchinr=1 kchinr=2 kchinr=3 kchinr=4 kchinr=5
Upper income group
0 5 5 11 19
Middle income group
3 15 7 6 7
Low income group 4 24 2 3 5
Indexes of de facto Financial Openness (International Financial Integration)
• Data source: Lane and Milesi-Ferretti’s "External Wealth of Nations" Dataset 1970-2007 and calculation based on the IMF IFS dataset • GFA: gross foreign assets
• GFL: gross foreign liabilities
• Indexes: • IFI1=100*(GFA+GFL)/PPPGDP
• IFI2=100*(GFA+GFL)/GDP
• IFI3=100*(GFA+GFL)/TRADE
-1.00
-0.50
0.00
0.50
1.00
1.50
1970 1975 1980 1985 1990 1995 2000 2005 20101A. Chinn-Ito Index
0
100
200
300
400
1970 1975 1980 1985 1990 1995 2000 2005 20101B. IFI2=100*(GFA+GFL)/GDP
0
100
200
300
1970 1975 1980 1985 1990 1995 2000 2005 20101C. IFI1=100*(GFA+GFL)/PPPGDP
50
100
150
200
250
300
1970 1975 1980 1985 1990 1995 2000 2005 20101D. IFI3=100*(GFA+GFL)/TRADE
Average Trend of International Financial IntegrationUpper income group Middle income group Low income group
Identifying “Currency Crash” Events
• Criterions for indentifying “currency crash” events ( a la Frankel and Rose 1996):
– Annual devaluation rate at least 25% – Devaluation rate is at least 10% higher than in the previous
year. – Events occurring within 3 years of an earlier event are
windowed out.
• 197 independent currency crash events are identified – Reduced to 175 events if we count the group devaluations by
the 13 CFA countries in 1981 or 1994 as a single event. – The magnitude of the devaluations rates for these events: median 59%; 25th percentile 37%; 75th 110%
6
2
3 3
1
5
3
1
8
6
8 8
6
5
7
5
2
10
9
6 6
10
5
3
6
11
7
2 2
6
5
1 1
7
1970 1974 1978 1982 1986 1990 1994 1998 2002 2006Year
The Frequency of Currency Crash Events in LDCs
46
64
25
11
6
1 2 3 4 5The Chinn-Ito Rating Index of Financial Openess
Distribution of Crises across de jure openness
0
10
20
30
40
50
10 20 30 40 50 60 70 80 90%
M2/Intl reserves minus gold
0
2
4
6
8
10
10 20 30 40 50 60 70 80 90%
Short-term debt/intl reserves
Behavior of indicators during the crises episodesPercentile curves
T-3:T T+1:T+3 control group
0
50
100
150
10 20 30 40 50 60 70 80 90%
Domestic credit growth
0
50
100
10 20 30 40 50 60 70 80 90%
Domestic credit to private sector growth
0
50
100
10 20 30 40 50 60 70 80 90%
M2 growth rate
-6
-4
-2
0
2
4
10 20 30 40 50 60 70 80 90%
budget balance(%GDP)
Behavior of indicators during the crises episodesPercentile curves
0
50
100
150
10 20 30 40 50 60 70 80 90%
Inflation rate
10
15
20
25
30
35
10 20 30 40 50 60 70 80 90%
Investment rate
-5
0
5
10
10 20 30 40 50 60 70 80 90%
GDP growth
-15
-10
-5
0
5
10 20 30 40 50 60 70 80 90%
Current account(%GDP)
Behavior of indicators during the crises episodesPercentile curves
0
10
20
30
40
50
10 20 30 40 50 60 70 80 90%
Ratio of debt liabilities to equity liabilities
0
50
100
150
10 20 30 40 50 60 70 80 90%
External debt(%GDP)
0
5
10
15
20
10 20 30 40 50 60 70 80 90%
Short-term debt(%GDP)
0
20
40
60
80
100
10 20 30 40 50 60 70 80 90%
Short-term debt(%Export)
Behavior of indicators during the crises episodesPercentile curves
Nonparametric Tests: (T-3:T) vs. Control Group
Variables Kruskal-Wallis
Kolmogorov-Smirnov t-test
M2/intl reserves 0 0 0.054
Short debt/intl reserves 0 0 0.009
Domestic credit growth 0 0 0.828
DC to private sector growth 0 0 0
DC to private sector/M2 0 0 0.002
Bank loans(% GDP) 0.392 0.312 0.82
M2 growth 0 0 0
Budget balance(% GDP) 0.001 0.011 0.005
Government debt(% GDP) 0.132 0.048 0.028
Inflation 0 0 0
Investment(% GDP) 0 0 0
Investment growth 0 0 0
Variables Kruskal-Wallis
Kolmogorov-Smirnov t-test
Consumption growth 0 0 0
GDP growth 0 0 0
Current account(%GDP) 0 0 0
Export growth 0 0 0
openness 0 0 0 de facto International Financial Integration1 0 0 0
de facto International Financial Integration2 0 0 0
Total debt(%GDP) 0 0 0
Short-term debt(%GDP) 0 0 0
Short-term debt/export & income 0 0 0
Short-term debt(% export) 0 0 0
Debt/equity ratio 0 0 0.079
Probit Estimates
• The model:
– K -- Chinn-Ito index
– X -- control variables
– In order to avoid the endogeneity problem and to smooth out the short-run fluctuations of the variables, we use averages of the T-3, T-2, T-1 values for all the regressors.
β)X KΘ(1)sProb(crisi iitit +== α
(1) (3) (5) (7) (10)VARIABLES ev ev ev ev ev
kchinl -0.0141** -0.0143** -0.0157*** -0.0144** -0.0114**(-2.525) (-2.503) (-2.649) (-2.520) (-2.064)
ifidl -0.000203** -0.000245** -0.000246** -0.000240**(-2.091) (-2.268) (-2.362) (-2.488)
imresl 0.000149** 0.000153** 0.000149**(1.984) (2.010) (1.981)
dc3gl 6.56e-06 -6.83e-06 -1.85e-05*(0.612) (-0.588) (-1.703)
drerl -0.000482** -0.000515** -0.000622** -0.000509** -0.000544**(-2.165) (-2.085) (-2.528) (-2.054) (-2.148)
sdx1l 0.000248*** 0.000220*** 0.000246*** 0.000240***(3.521) (3.406) (3.509) (3.534)
ygl -0.00687*** -0.00653*** -0.00669*** -0.00657*** -0.00595***(-3.962) (-3.615) (-3.646) (-3.602) (-3.347)
cayl -0.00256** -0.00213** -0.00173* -0.00214** -0.00122(-2.414) (-1.961) (-1.741) (-1.975) (-1.105)
sdrl 0.000145**(2.072)
ifipl -0.000234(-1.327)
dcpv3gl -6.33e-06(-0.671)
m2gl 5.90e-06(0.348)
gresml -0.00646**(-2.338)
dertol
gvcgl
sdxl
sdtdl
Observations 1779 1735 1728 1734 1745Pseudo R-squared 0.0626 0.0741 0.0692 0.0742 0.0767Robust z-statistics *** p<0.01, ** p<0.0
kchinl ifidl sdrl dc3gl drerl ygl cayl-0.018 -0.021 0.008 0.003 -0.015 -0.024 -0.015
kchinl ifidl imresl dcpv3gl drerl sdx1l ygl cayl -0.018 -0.026 0.008 -0.003 -0.016 0.016 -0.023 -0.013
Marginal effect on the likelihood if the regressor goes up by one standard deviation
• Given that all other regressors are evaluated at the mean, the likelihood that a country encounters a crisis will drop to 0.032 from 0.089 if it moves from the lowest degree of capital account openness to highest degree of openness.
– Marginal effects after dprobit – y = Pr(ev) (predict, p) – = .03225865
– Marginal effects after dprobit – y = Pr(ev) (predict, p) – = .0894081
• Finding: currency crash events are more likely to occur
under a setting of tighter capital controls.
• Why?
• “Core” and “Peripheral” research themes in the literature
“The Costs and Effects of Capital Controls – A Review of Recent Literature and Implications for China” (working paper, Huang 2012)
– Core: Does financial openness promote higher economic growth?
– Peripherals:
1. Are capital controls effective in preventing large volumes of capital inflows or outflows?
2. Will capital controls bring about monetary autonomy for the countries who aims to stabilize their exchange rates?
3. Do capital controls bring about distortion and corruption to the economy?
• “Core” and “Peripheral” research themes in the literature – Core: Does financial openness promote higher
economic growth?
– Peripherals:
4. Will capital account liberalization promote the development of domestic financial system?
5. Does financial openness possess “discipline effects”– to discipline and motivate the governments to pursue macro-prudential and sustainable policies?
6. Are capital controls effective in preventing financial crises? Or do capital controls help the economies to fare better in the aftermath of the financial crises?
Why capital control is linked to higher odds of currency crises?
• Theoretical arguments against capital controls: – From microeconomic perspective: capital control can
cause distortion and facilitate corruption with the cost of reduced efficiency and resource mis-allocation, resulting in lower productivity and growth potential.
• Rajan and Zingales (1998) • Johnson and Mitton (2002), Desai et. al. (2002), Forbes (2003)
– From macroeconomic perspective, financial openness may possess a “discipline effect”. Therefore, policy makers may be subject to adverse selection or moral hazard problems when deciding capital account policies.
• Obstfeld (1998),Tytell and Wei (2005)
Why capital control is linked to higher odds of currency crises?
• Empirical exploration about the relationship
between economic performance and capital controls:
kchinr vs. economic fundamentals
Economic fundamentals vs. Chinn-Ito Rating Index kchinr
Variables 1 2 3 4 5
M2/non gold reserves 8.149 4.003 3.556 3.275 3.167
Domestic credit growth 21.717 16.763 16.216 15.246 13.569 DC to private sector growth 24.459 17.425 18.478 18.589 15.777
DC to private sector(%M2) 14.311 10.616 12.604 12.100 10.872
M2 growth 24.244 17.234 17.687 16.402 14.068
Budget balance(%GDP) -0.841 -2.137 -1.675 -1.525 -0.896
Central govt debt(%GDP) 26.873 59.416 52.921 42.927 39.842
inflation 17.034 8.191 8.934 6.598 4.154
Investment (%GDP) 19.965 20.600 22.657 21.576 23.898
variables 1 2 3 4 5
Investment growth 4.999 5.622 6.057 6.966 7.838
Consumption growth 3.251 4.005 4.463 4.105 5.745
GDP growth 4.000 4.400 4.755 4.253 5.668
Export growth 4.800 5.955 6.695 6.700 8.303
openness 46.176 56.063 68.769 68.036 96.161
De facto k open1 68.200 94.600 96.400 105.400 194.500
De facto k open2 34.000 41.100 48.500 49.900 107.400
Debt/equity ratio 5.460 4.314 2.812 2.183 2.730
Why capital control is linked to higher odds of currency crises?
• Capital control practices are not effective in limiting capital inflows or outflows. – Edwards(1999,2001),Simone and Sorsa (1999),
Valdés and De Gregorio (2000),Cowan and De Gregorio (2005) conclude that the “encaje” policy didn’t effectively reduce the total volume of capital inflows, but only had limited effect in lengthening the maturity of capital inflows.
– Garcia and Barcinski (1998),Garcia and Valpassos (2000),Goldfajn and Minella (2005),Carvalho and Garcia (2006) come to similar conclusions regarding Brazil’s capital control practices.
Conclusions and Policy Implications
• In general, capital control does not reduce the probability of financial crises. On the contrary, the tighter the control, the more likely the inertia to postpone reforms, hence leading to destablizing elements, and resulting in a higher likelihood of financial crises.
• Under the ongoing trend of global financial integration, capital control measures are becoming increasingly ineffective while incurring increasingly higher costs and distortions.
• A wiser approach to strenghen financial safety is to push on necessary reforms and institutional development, to pursue sustainable macroeconomic policies and to improve the macroeconomic fundamentals.