UNIVERSITY OF CALIFORNIA SANTA CRUZ ESSAYS ON EMPIRICAL ANALYSIS OF CAPITAL CONTROLS A dissertation submitted in partial satisfaction of the requirements for the degree of DOCTOR OF PHILOSOPHY in ECONOMICS by Jongchan Lee December 2016 The Dissertation of Jongchan Lee is approved: Professor Nirvikar Singh, Chair Professor Carl E. Walsh Professor Eric M. Aldrich Tyrus Miller Vice Provost and Dean of Graduate Studies Copyright © by Jongchan Lee 2016 Table of Contents List of Figures v List of Tables vi Abstract viii 1 Did Pre-crisis Capital Controls Mitigate the Severity of the Global Financial Crisis? - Evidence from Advanced, Emerging, and De- veloping countries 1 1.2 Brief history of the debate on financial liberalization .1 Financial liberalization and economic stability risk .2 Capital controls and GDP collapse in the Global Financial Crisis .4 Data and methodology .2 Capital controls data .6 New measure of crisis severity - GDP loss rate .7 Discussion of results. 34 2 Effectiveness of Precautionary Capital Controls - Generalized Propensity Score Approach 57 2.3 Generalized propensity score analysis .2 Generalized propensity score (GPS) analysis .3 GPS - simple example with visualization .4 GPS and our study .3 Covariates - Determinants of capital controls .1 Pre-treatment covariates and estimated propensity score .2 Evidence of the balancing property of GPS .3 Dose-response function.
86 3 Financial Liberalization, Financial Crisis, and Consumption in Korea 110 3.3 Korea’s financial globalization in the 2000s and the sudden stop episode in 2008-2009 .1 The Asian Financial Crisis .2 Financial liberalization policy in the inter-crisis period .3 External debt expansion in the run-up to the Global Finan- cial Crisis .4 Sudden stop in Korea in 2008-2009 .4 Financial liberalization and consumption .5 Effect of easing inflow capital controls on consumption growth .6 Effect of financial globalization on consumption development in 2005-2010 .1 Synthetic control matching. 138 Bibliography 147 iv List of Figures 1.1 Crisis severity - Demeaned GDP growth rate and GDP forecast error 37 1.2 Demeaned GDP growth rate and log GDP per capita .3 De jure capital controls from 1995 to 2013 .4 De jure capital controls from 1995 to 2013 - Annual volatility .5 GDP loss rate .1 Example : Estimated propensity score .2 Example : Estimated propensity score .3 Example : Estimated propensity score .4 Capital controls and GDP per capita .5 Capital controls and demeaned growth .6 Capital controls and GDP forecast error .7 Capital controls and GDP difference .8 Dose-response function : Inflow capital controls and demeaned growth 95 2.9 Dose-response function : Inflow capital controls and GDP forecast error .10 Dose-response function : Outflow capital controls and GDP forecast error .1 De jure capital liberalization in Korea .2 External debt in Korea .3 Arbitrage opportunity of Korean financial asset .4 Result of synthetic control matching - Consumption growth. 144 v List of Tables 1.1 Crisis severity measure .2 Country group categorization .3 Data statistics of capital controls .5 Demeaned growth rate and inflow capital controls in all countries 46 1.6 Demeaned growth rate and inflow capital controls in advanced countries .7 Demeaned growth rate and inflow capital controls in emerging coun- tries .8 Demeaned growth rate and inflow capital controls in developing countries .9 GDP forecast error and inflow capital controls in all countries .10 GDP forecast error and inflow capital controls in advanced countries 51 1.11 GDP forecast error and inflow capital controls in emerging countries 52 1.12 GDP forecast error and inflow capital controls in developing countries 53 1.13 Demeaned growth rate and outflow capital controls in advanced countries .14 GDP forecast error and outflow capital controls in advanced countries 55 1.15 GDP forecast error and debt inflow capital controls in developing countries .1 Crisis severity measure .2 Correlations between crisis severity measures .3 Percentage of countries which changed capital controls .4 Pre-crisis inflow/outflow capital controls .5 Potential determinants of capital controls .6 Determinants of inflow capital controls .7 Determinants of inflow capital controls .8 Estimated generalized propensity score .9 Balancing property of GPS .10 Expected conditional outcome given treatement and GPS .11 Expected conditional outcome given treatement and GPS .12 Expected conditional outcome given treatement and GPS .1 IV regression results of consumption growth .2 Country weights of the synthetic control country. 146 vii Abstract Essays on empirical analysis of capital controls by Jongchan Lee The Global Financial Crisis (GFC) of 2008-09 focused attention on the possibility that pre-crisis capital controls had beneficial effects on post-crisis GDP perfor- mance.
The GFC provides a good case for such analysis, because it was almost unexpected and its impact was worldwide. The first chapter revisits this issue and makes two main contributions. First, I improve on the methodology of pre- vious studies in our choices of control variables, measurement of capital controls, estimating timing of pre-crisis capital controls, and allowing for different types of capital flows. Second, I apply a new measure of crisis severity and apply it to the case of the impact of capital controls on post-crisis economic performance.
Findings from cross-sectional regression show that in advanced countries a higher degree of pre-crisis outflow controls are associated with less severe GDP collapse. In emerging countries, any type of pre-crisis capital controls is not found signifi- cant. For developing economies, I found only a partial result that stricter pre-crisis debt inflow capital controls were associated with an even larger GDP collapse than in the absence of such controls. The second chapter investigates the effects of ex ante capital controls in mit- igating the output cost of the 2008-2009 Global Financial Crisis (GFC).
Restric- viii tions on capital accounts are accepted as a tool for ensuring financial stability among the policy community after the GFC, but the empirical evidence of capital controls’ usefulness has not yet been clearly verified. I employ Imbens (2000)’s generalized propensity score (GPS) analysis using 88 cross-country data to handle this issue. GPS analysis is designed to find the average treatment effect where the treatment is continuous. The GPS approach in combination with the balanc- ing property can remove all biases from country characteristics related to capital controls.
I found indeterminate evidence that pre-crisis inflow capital controls successfully lessened the GDP cost of the GFC. In addition, I found a partial re- sult that pre-crisis outflow capital controls may even worsen GFC crisis severity. A contribution of the second chapter is that the GPS method limits the endo- geneity of capital controls. Another contribution is that I detect non-linear effect of capital controls on the economic stability.
According to "overborrowing" theory, financial globalization and the subse- quent rise in consumption sourced by foreign debt in normal times can worsen the impact of financial crises. However, the theory of international risk sharing predicts that financial liberalization helps stabilize consumption behaviors. A contribution of the third chapter is that I investigate the effect of de jure finan- cial liberalization in Korea on consumption growth in the context of the Global Financial Crisis using novel empirical methods. Another contribution of the third chapter is that I document Korea’s de jure and de facto financial globalization ix episode in the 2000s and the sudden stop experience during the Global Finan- cial Crisis (GFC) comprehensively.
I found from IV regression that the economic effect of easing capital inflow controls in Korea was large for the acceleration of consumption growth, but this effect was statistically insignificant. Limited evi- dence from synthetic control matching shows that financial liberalization in Korea did not significantly affect consumption collapse during the GFC. x Chapter 1 Did Pre-crisis Capital Controls Mitigate the Severity of the Global Financial Crisis? - Evidence from Advanced, Emerging, and Developing 1 countries 1.1 Introduction Whether financial liberalization is adversely associated with financial stabil- ity has been a topic of hot debate for decades. Various channels through which financial liberalization causes or amplifies external crisis are considered.
Sudden contraction of capital flow raises market interest rate sharply, and investment and durable goods consumption collapse. Real estate asset price drops as well through the financial channel. Countries exposed with large external financing face with liquidity problems. Depreciation by capital flight and pessimistic expectation lowers the US-dollar value of domestic assets, which are implicitly or explicitly considered as collateral for international borrowing.
The decreased value in col- lateral causes more capital outflow. In addition, the value of foreign assets held by residents are vulnerable to global shock. After the Asian Financial Crisis 1997-98, this topic was studied by empirical researchers rigorously since the crisis coincided with rapidly opened capital ac- counts in Asia. However, empirical literature could not find strong cross-country evidence that financial liberalization is associated with higher incidence or inten- sity of external crisis.
In addition, among the empirical literature which studies the effect of capital controls on capital flow, there is no consensus that capital 2 controls actually affect the capital flow. In the event of the Global Financial Crisis 2008-2009, most advanced and emerging countries experienced output collapse. The Global Financial Crisis is a good context in which to investigate the effect of capital controls on the out- put collapse or crisis severity, because it was almost an unexpected event and its impact was worldwide. Since each country has its own degree of financial lib- eralization, studying whether the degree of financial openness affects the output collapse after the worldwide shock is a quasi-experimental research design.
There is empirical literature on the impact of pre-crisis capital controls on post-crisis economic performance, including papers such as Ostry et al. (2012) and Blundell-Wignall and Roulet (2013): this literature is reviewed in the next section. The present paper makes two main contributions. First, we improve on the methodology of previous studies in our choices of control variables, measure- ment of capital controls, estimating the timing of capital controls, and allowing for different types of capital flows.
Second, we apply a new measure of crisis severity and apply it to the case of the impact of capital controls on post-crisis economic performance. We improved the methodology of previous research in a comprehensive way and found different results. We employed various crisis severity measures to guarantee robustness of the results. We changed the timing of capital controls reasonably and investigated the effect of various types of capital controls by direction of flow and 3 by asset category, which are inflow1 capital controls, outflow capital controls, debt capital controls, and equity capital controls.
In addition, we selected the control variables more carefully. Domestic credit to GDP, house real price appreciation, GDP per capita, and exchange rate regime were added as control variables since they had been found significant in previous studies on the severity of the Global Financial Crisis. We investigated all groups of countries with advanced, emerging, and developing economies. Another innovation of this research is that we develop a new measure of crisis severity : the GDP loss rate.
The new measure is somewhat similar to other measures of GDP loss, but has never been formulated in precisely this way and never applied in this manner. It measures decreased percentage of the actual GDP relative to the counterfactual GDP path. The GDP loss rate has several advan- tages over the previous measure. First, it reveals the concept of the counterfactual GDP path explicitly.
Second, what it measures is conceptually clearer than other measures, how much percentage the actual GDP decreased from a hypothetical GDP path. Third, application of this measure is flexible according to various definitions of the counterfactual GDP path. Our findings are as follow. For advanced countries, higher degree of pre-crisis 1 The definition of capital inflow in this paper is non-resident’s purchase of domestic assets or resident’s sale of foreign assets.
Inversely, capital outflow is defined as resident’s purchase of foreign asset or non-resident’s sale of domestic asset. This directional definition of capital inflow/outflow is different from the residence-based definition, which is non-resident’s purchase or sale of domestic assets as capital inflow and resident’s purchase or sale of foreign assets as capital outflow. 4 outflow controls are associated with less severe GDP collapse.