1. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
Is financial globalization welfare decreasing?
Marcel Schr¨der
o
Panel: Prema-chandra Athukorala, Paul Burke, Ippei Fujiwara
Crawford School PhD Conference
November 27, 2012
2. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
Some Background
• For any open economy, net foreign asset position (NFA) is a
key variable: It limits present value of future current account
deficits.
• Net external asset position represents a country’s solvency
constraint.
• Traditional view: ∆N F At = CAt .
• But: Foreign assets and liabilities are measured at market
value.
• Market value variations occur due to changes in asset prices,
asset returns, or exchange rate changes → ”Valuation
effects”.
3. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
• Valuation effects are an important driver of NFA (Lane and
Milesi-Ferreti, 2007; Gourinchas, 2007; Obstfeld, 2004):
∆N F At = CAt + V ALt . (1)
• However, capital transfers (CAP) and unrecorded capital
flows/trade flows are also important:
∆N F At = CAt + V ALt + CAPt + EOMt . (2)
• Can compute V ALt indirectly using Eq. 2.
• Magnitude of VAL is proportional to gross asset positions.
• Proliferation in asset trade (financial globalization) has led to
a significant increase in the size of valuation adjustments.
4. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
Increasing volatility in valuation effects over time
Figure: Valuation-Effect Volatility and Financial Integration, 1980-2007
High Income
.1
6
5
Valuation−Effect Volatility
.08
(A+L)/GDP
4
.06
3
.04
2
.02
1
1980 1990 2000 2010
Year
VAL Financial_Integration
Note: Volatility measured as the rolling standard deviation over 10-year periods.
Source: Compiled from Lane and Milesi-Ferretti (2007) and IMF BOP statistics.
5. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
Figure: Valuation-Effect Volatility and Financial Integration, 1980-2007
Emerging
.07
3
Valuation−Effect Volatility
2.5
.06
(A+L)/GDP
2
.05
1.5
.04
1
.03
1980 1990 2000 2010
Year
VAL Financial_Integration
Note: Volatility measured as the rolling standard deviation over 10-year periods.
Source: Compiled from Lane and Milesi-Ferretti (2007) and IMF BOP statistics.
6. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
Figure: Valuation-Effect Volatility and Financial Integration, 1980-2007
Developing
1.8
.1
1.6
Valuation−Effect Volatility
.08
1.2 1.4
(A+L)/GDP
.04 .06
1
.8
.02
1980 1990 2000 2010
Year
VAL Financial_Integration
Note: Volatility measured as the rolling standard deviation over 10-year periods.
Source: Compiled from Lane and Milesi-Ferretti (2007) and IMF BOP statistics.
7. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
Research Question
• How does the rising importance of VEs affect economic
performance?
• Empirically, this remains an open question.
• The purpose of the study is to fill this gap.
• Here, focus on welfare through consumption volatility.
• Currently, other channels not well understood theoretically.
8. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
Theoretical considerations
• Theoretical link between VEs and consumption is not
clear-cut.
• Interpretation 1: Higher valuation-effect volatility causes
greater volatility in wealth ⇒ consumption should become
more volatile as well.
• For risk averse agents this means a welfare loss stemming
from deviations from a smooth consumption path.
• Interpretation 2: VEs reflect flow payments of international
risk sharing → wealth not more volatile → no welfare costs.
9. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
What to expect?
• The relationship between consumption and valuation-effect
variability is conditional on:
• The currency decomposition of a country’s balance sheet.
• And/or its ability to share risk internationally.
• Since Eichengreen & Hausmann (1999) it is well known that
most developing countries face problem of ”original sin”.
• Kose et al (2009): Developing economies still unable to share
risk internationally despite financial globalization.
10. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
Empirical model
LV OLCit = β0 +β1 LV OLV ALit +β2 LV OLGit +β3 LV OLIN F Lit +u† .
it
• LV OLC: Volatility of real private consumption growth per capita.
• LV OLV AL: Volatility of valuation effects.
• LV OLG: Volatility of real GDP per capita growth.
• LV OLIN F L: Volatility of inflation rate.
• Volatility defined as log of rolling standard deviation using 10yr windows.
• β1 insignificant: cannot reject that VEs reflect risk sharing.
• β1 positive significant: VE-volatility welfare decreasing, reject
risk sharing interpretation.
11. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
Data
• Unbalanced panel, 82 countries. Period: 1980-2007.
• Data Sources:
• LV OLC & LV OLG: PWT 7.1.
• LV OLIN F L: WDI.
• LV OLV AL: Compiled from: EWN II (Lane and
Milesi-Ferretti, 2007) and IMF BOP statistics.
12. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
Estimation and results
Method
• Step 1: Investigate unit root properties of the variables.
• Step 2: Estimate long-run relationship between variables.
• Step 3: Test for cointegration.
13. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
Estimation: Dynamic OLS (Mark and Sul, 2003)
LV OLCit = αi + θt + λi t + β xit + u† ,
it
m
u† =
it δi ∆xi,t+m + uit .
−m
• xit = [LV OLV ALit , LV OLGit , LV OLIN F Lit ]
• Assumptions:
• Homogenous cointegrating vector, [1, −β ].
• u† independent across countries.
it
• Group-specific heterogeneity accounted for by:
Fixed effects, time fixed effects, heterogeneous time trends,
disparate short-run dynamics.
14. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
Results
Table: Baseline Regression Results
(1) (2) (3) (4)
Independent variable Full Sample High Income Emerging Developing
Valuation-Effect Volatility 0.085 (0.025)∗∗∗ -0.045 (0.050) 0.162 (0.052)∗∗∗ 0.082 (0.029)∗∗∗
Real GDP Growth Volatility 0.677 (0.036)∗∗∗ 0.978 (0.066)∗∗∗ 0.811 (0.086)∗∗∗ 0.458 (0.037)∗∗∗
Inflation Volatility 0.103 (0.013)∗∗∗ -0.009 (0.042) 0.093 (0.02)∗∗∗ 0.129 (0.021)∗∗∗
Country Specific Effects Yes Yes Yes Yes
Heterogeneous Time Trends Yes Yes Yes Yes
Observations 1629 423 391 815
N 82 19 20 43
IPS (p-value) 0.00 0.00 0.00 0.00
Notes: ∗∗∗ , ∗∗ , ∗ denote the level of statistical significance at 1, 5, and 10 percent. Standard errors in parentheses.
IPS refers to the Im et al. (2003) unit root test performed on the residuals under the unit root null.
15. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
• PBM predicts: Impact of VE-volatility on consumption
volatility depends on currency decomposition of external debt.
• I construct subsamples based on the foreign currency exposure
of countries’ external debt.
• I consider the following measure:
• Weight of debt liabilities (portfolio debt and other debt)
denominated in foreign currency.
• Source: Lane and Shambaugh (2010).
16. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
Table: Results for subsamples based on foreign currency exposure
(1) (2) (3)
Independent variable DL100 DL<100 DL100
Non High Income Emerging
Valuation-Effect Volatility 0.116 (0.031)∗∗∗ 0.117(0.053)∗∗ 0.236 (0.097)∗∗
Real GDP Growth Volatility 0.536 (0.040)∗∗∗ 0.707 (0.086)∗∗∗ 1.042 (0.16)∗∗∗
Inflation Volatility 0.114 (0.015)∗∗∗ 0.103 (0.026)∗∗∗ 0.097 (0.027)∗∗∗
Country Specific Effects Yes Yes Yes
Heterogeneous Time Trends Yes Yes Yes
Observations 1001 205 204
N 52 11 10
IPS (p-value) 0.00 0.00 0.00
Notes: ∗∗∗ , ∗∗ , ∗ denote the level of statistical significance at 1, 5, and 10 percent. Standard
errors in parentheses. IPS refers to the Im et al. (2003) unit root test performed on the residuals
under the unit root null.
17. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
Conclusion
• The results are consistent with a priori theoretical predictions.
• They suggest that valuation-effect volatility imposes welfare
costs on countries with a high share of their liabilities
denominated in foreign currency.
• Implication of the findings is that financial globalization
decreases welfare in emerging and developing economies that
face the problem of original sin (partial equilibrium).
• These types of countries should take a cautious approach to
liberalizing their capital account transactions.
18. Background Research Question Theoretical considerations Empirical model and data Estimation and results Conclusion
Thank You!