EXCHANGE RATE VOLATILITYEXCHANGE RATE VOLATILITYTATRE JANTARAKOLICA, PORJAI CHALERMSOOK, AND PALAKORN SATSUE
PRESENTED BY
TATRE JANTARAKOLICAPORJAI CHALERMSOOK
At Thammasart University
Introduction
G l T h h t l tilit ff t Goal: To see how exchange rate volatility affects the export
Use the estimated results to support policy making
Exchange Rate Volatilityg y
Exchange Rate Volatility = Exchange Rate Volatility Risk associated with unexpected movement in exchange ratesg
No precise definition on how to measure exchange No precise definition on how to measure exchange rate volatility
Literature and History
Historyy
The breakdown of Bretton-Woods Agreement The breakdown of Bretton Woods Agreement allows floating exchange rate
Common Belief: Weakness of fl ti h t ifloating exchange rate regime
High Exchange Rate Volatility High Exchange Rate Volatility
Higher cost for risk-averse traders
Lower Trade
Common Belief: Weakness of fl ti h t ifloating exchange rate regime
High Exchange Rate Volatility High Exchange Rate Volatility
Higher cost for risk-averse traders
Lower Trade
So floating exchange rate g gregime is believed to discourage Trade!!!
Common Belief: Weakness of fl ti h t ifloating exchange rate regime
High Exchange Rate Volatility High Exchange Rate Volatility
Higher cost for risk-averse traders
Lower Trade
So floating exchange rate EURO currency is used partly g gregime is believed to discourage Trade!!!
y p yto deal with volatility
Historyy
Belief: Trade suffers from volatility of exchange Belief: Trade suffers from volatility of exchange rates
Research (both theoretical and empirical) has failedto conclusively support this hypothesis!! to conclusively support this hypothesis!!
Previous work
Use various measures of volatility and different Use various measures of volatility and different modelsPossible relationships between trade and volatility Possible relationships between trade and volatility
Positive Negative No Effect
[McKenzie and Brooks [Akhtar and Hilton 84], [McKenzie 99]97]
,[Arize 95], [Arize 96]
[Arize 98], [Arize and Malindretos 98]
[Arize and Ghosh 94], [Lastrapes and Koray 90]
[Arize 98], [Arize and Shwiff 98]
[Peree and Steinherr 89] [Asseery and Peel 91] [Bahmani-Oskooee[Peree and Steinherr 89] [Asseery and Peel 91], [Bahmani-Oskooee 02]
[Bahmani Oskooeeand Payesteh 93]
[Corbo and Caballero 89], [Bailey et al 87]
[Bailey et al 87]
[Kl 04] [H ll 95] [D Vit d Abb tt [Klassen 04], [Holly 95],[Kroner and Lastrapes 93]
[De Vita and Abbott 04]
[Our work]
Previous work
Why are the results so different? Why are the results so different?
Previous work
Why are the results so different? Why are the results so different?
They are sensitive to: They are sensitive to: Proxies for Exchange Rate Volatility Model SpecificationModel SpecificationSample period Countries consideredCou es co s de ed
Previous work
Why are the results so different? Why are the results so different?
They are sensitive to: They are sensitive to: Proxies for Exchange Rate Volatility Model Specification Central Issue!Model SpecificationSample period Countries considered
Central Issue!
Cou es co s de ed
Remark
What is the rationale for positive relation? What is the rationale for positive relation?
Remark
What is the rationale for positive relation? What is the rationale for positive relation?
[De Grauwe 88] “The effect depends on the [De Grauwe 88] The effect depends on the degree of risk aversion and can be either positive or negative”or negative
Remark
What is the rationale for positive relation? What is the rationale for positive relation?
[De Grauwe 88] “The effect depends on the [De Grauwe 88] The effect depends on the degree of risk aversion and can be either positive or negative”or negative
Very risk-averse
Worry about worst possible outcomeWorry about worst possible outcome
Remark
What is the rationale for positive relation? What is the rationale for positive relation?
[De Grauwe 88] “The effect depends on the [De Grauwe 88] The effect depends on the degree of risk aversion and can be either positive or negative”or negative
Very risk-averse
Worry about worst possible outcomeWorry about worst possible outcome
Export more to avoid huge loss
Remark
What is the rationale for positive relation? What is the rationale for positive relation?
[De Grauwe 88] “The effect depends on the [De Grauwe 88] The effect depends on the degree of risk aversion and can be either positive or negative”or negative
Very risk-averse
Worry about worst possible outcomeWorry about worst possible outcome
Export more to avoid huge loss
Export increases
Remark
[De Grauwe 88] “The effect depends on the [De Grauwe 88] The effect depends on the degree of risk aversion and can be either positive or negative” g
Very risk-averse Less risk-averse
Worry about worst possible outcome
y
Find the revenue less attractive
Export more to avoid huge loss Export less
Export increases
Related research questions and R t d l tRecent developments
[Freund-Pierola 08] [Freund Pierola 08]
Export surges = significant increase in export growth and lasts for at least 7 yearsgrowth and lasts for at least 7 years
Results: Results: Real depreciation+ volatility reduction export surge (in developing countries) (in developing countries)
Related research questions and R t d l tRecent developments
[Alvarez-Doyle-Lopez 09] started a new direction [Alvarez Doyle Lopez 09] started a new direction (Intensive margin) Export quantity of a fixed product (Extensive margin) Range of goods exported(Extensive margin) Range of goods exported
US Trade dataUS Trade dataResults:
No effect on intensive marginNo effect on intensive marginSignificant negative effect on extensive margin
Proxies for Volatility
Proxies for Volatilityy
Standard theory of International EconomicsStandard theory of International Economics
ttt Xe εβα ++=
Spot exchange rate
Proxies for Volatilityy
Standard theory of International EconomicsStandard theory of International Economics
ttt Xe εβα ++=
Spot exchange rate
Explanatory variable
Proxies for Volatilityy
Standard theory of International Economics
ttt Xe εβα ++=
Standard theory of International Economics
Spot exchange rate
Explanatory variable
Error term
Proxies for Volatilityy
Standard theory of International Economics
ttt Xe εβα ++=
Standard theory of International Economics
Some candidates for measuring volatilityV i f Variance of et
Variance of f
tε
Variance of 1−− tt εε
Summary of Previous worky
Measure Papers
Absolute percentage change of et [Thursby and Thursby 98][Bailey et al 86]
Moving average [Cushman 83] [Cushman 86][Akhtar and Spence-Hilton 84][Gotur 85][Kenen and Rodrik 86] [Klassen 04][Klassen 04]
ARIMA model residuals [Asseery and Peel 87][McIvor 95]
Standard deviation of the yearly percent [De Grauwe and Bellefroid 86]Standard deviation of the yearly percent around the mean age changes of exchange rate
[De Grauwe and Bellefroid 86][De Grauwe 87] [De Grauwe 88]
ARCH models [Pozo 92] [Kroner and Lastrapes 93] ARCH models [Pozo 92] [Kroner and Lastrapes 93] [McKenzie and Brooks 97] [McKenzie 98][Qian and Varangis 94]
Proxies for Volatility: Other Issuesy
Real or Nominal Exchange Rates? Real or Nominal Exchange Rates?
Proxies for Volatility: Other Issuesy
Real or Nominal Exchange Rates? Real or Nominal Exchange Rates? Again … this has been a long debate …
Proxies for Volatility: Other Issuesy
Real or Nominal Exchange Rates? Real or Nominal Exchange Rates? Again … this has been a long debate …Early work: Early work:
[Ethier 73], [Clark 73], [Baron 76] [Hooper and Kohlhagen 78], [Bini-Smaghi 91] [Akhtar and Spence-Hilton 84]78], [Bini Smaghi 91] [Akhtar and Spence Hilton 84]
• Risk (=Volatility) should be measured independent of price movementprice movement
• Nominal exchange rate was preferred
Proxies for Volatility: Other Issuesy
[Gotur 85] and [Cushman 83] criticized the nominal [Gotur 85] and [Cushman 83] criticized the nominal exchange rate and suggested the use of real exchange rateg
Proxies for Volatility: Other Issuesy
[Gotur 85] and [Cushman 83] criticized the nominal [Gotur 85] and [Cushman 83] criticized the nominal exchange rate and suggested the use of real exchange rateg
IMF research team and [Chan and Wong 85] also gused real exchange rate
Proxies for Volatility: Other Issuesy
[Gotur 85] and [Cushman 83] criticized the nominal [Gotur 85] and [Cushman 83] criticized the nominal exchange rate and suggested the use of real exchange rateg
IMF research team and [Chan and Wong 85] also gused real exchange rate
[Qian and Varangis 94], [Thursby and Thursby 87] and some others“Real or Nominal does not matter!”
Proxies for Volatility: Other Issuesy
Summary Summary [empirical] The choice does not matter [theoretical] Real exchange rate is better [Clark et al [theoretical] Real exchange rate is better [Clark et al 04]
Our choice: Real Effective Exchange Rate
Our proxiesp
We use two methodsWe use two methodsTime-Rolling Moving Average VarianceGARCHGARCH
Our proxiesp
We use two methodsWe use two methodsTime-Rolling Moving Average VarianceGARCHGARCH
Time-Rolling Moving Average Variance= variance Time Rolling Moving Average Variance variance of the j previous time steps
( ) ( )21 j
V f f f∑( ) ( )1
11t t t i j
iV fx fx fx
j −=
= −− ∑
1 j
fx fx∑1
t iji
fx fxj −
=
= ∑
Our proxiesp
We use two methodsWe use two methodsTime-Rolling Moving Average VarianceGARCHGARCH
Time-Rolling Moving Average Variance= variance Time Rolling Moving Average Variance variance of the j previous time steps
( ) ( )21 j
V f f f∑ In our experiment:( ) ( )1
11t t t i j
i
V fx fx fxj −
=
= −− ∑
1 j
fx fx∑
In our experiment: j= 3, 6, 9, 12
1t ij
ifx fx
j −=
= ∑
Our proxiesp
GARCH(p,q)GARCH(p,q)
( ) ( )dt tL fx L uφ θ⋅Δ =
( ) ( )qLLLL θθθθ 211( ) ( )qqLLLL θθθθ −−−−= L211
( ) ( )ppLLLL φφφφ −−−−= L2
21
11
Our proxiesp
GARCH(p,q)GARCH(p,q)
( ) ( )dt tL fx L uφ θ⋅Δ =
( ) ( )qLLLL θθθθ 211( ) ( )qqLLLL θθθθ −−−−= L211
( ) ( )ppLLLL φφφφ −−−−= L2
21
11
Our work: GARCH(1,1)
The Trade Model
The Model
CategoriesCategoriesAggregateDisaggregated, bilateralDisaggregated, bilateralDisaggregated, sectoral data
The Model
CategoriesCategoriesAggregateDisaggregated, bilateral
This work
Disaggregated, bilateralDisaggregated, sectoral data
The Model
Assumptions: Assumptions: Our export depends on the
Prices Exchange ratesProductionImportForeign Direct InvestmentVolatilityVolatility
Sttstststststsst fxvolfdiimmpifxepiex εβββββββ +++++++= 6543210 Sttstststststsst ffpfp βββββββ 6543210
The Model
Add dummy variable to account for subprime crisisAdd dummy variable to account for subprime crisis
Scrisisfxvolfdiimmpifxepiex εββββββββ ++++++++= 76543210 Sttststststststsst crisisfxvolfdiimmpifxepiex εββββββββ ++++++++ 76543210
The Model
Add dummy variable to account for subprime crisisAdd dummy variable to account for subprime crisis
Scrisisfxvolfdiimmpifxepiex εββββββββ ++++++++= 76543210
Data
Sttststststststsst crisisfxvolfdiimmpifxepiex εββββββββ ++++++++ 76543210
Data
Monthly dataJan 2000 to March 2010Jan 2000 to March 2010
Estimation Techniquesq
Static model – OLS Static model OLS Dynamic model – VECM
Estimation Techniquesq
Static model – OLS Static model OLS Dynamic model – VECM
Non-stationarity and Cointegration of data have l b i ialso been very important issues
Results
UNIT ROOT TEST USING AUGMENTED
Variable t-stat (level I(0)) t-statistic at frist difference I(1)
UNIT ROOT TEST USING AUGMENTEDDICKEY-FULLER TEST
( ( )) ( )
Ex -1.013 -21.837 ***
Epi 0.975 -5.762 ***
Fx -1.616 -10.994 ***
Mpi -1.009 -24.825 ***
Im -1.076 -21.842 ***
Fdi -8.677 *** -24.246 ***
reerrisk3 -5.440 *** -15.426 ***
reerrisk6 -2.677 * -12.664 ***
reerrisk9 -1.982 -11.615 ***
reerrisk12 1 616 10 951 ***reerrisk12 -1.616 -10.951 ***
Reervol -2.161 -8.442 ***
Results: Static model
Static Model
(1) (2) (3) (4) (5) (6)
Constant -7590.6580 *** -7315.1830 *** -7699.0580 *** -8063.3320 *** -8641.4330 *** -7879.4330 ***
Epi 84.9185 *** 74.9425 *** 85.4443 *** 85.9543 *** 83.4508 *** 86.9656 ***
Mpi 47.9483 *** 49.9264 *** 47.8419 *** 47.9305 *** 50.1391 *** 46.8230 ***
Im 0.2362 *** 0.2500 *** 0.2332 *** 0.2384 *** 0.2414 *** 0.2312 ***
Fdi 0.1078 0.0710 0.1050 0.0856 0.0979 0.1087
Crisis -232.6986 -165.3934 -245.8626 -341.2409 -464.2319 -286.7422
Fx -91.6772 * -61.1945 -89.2590 * -86.5488 * -67.2174 -107.1289 **
reerrisk3 79.2395 ***
reerrisk6 33.3553
reerrisk9 109.5381 *
reerrisk12 226.8153 ***
Reervol 197.5235 **
Overall Test 179.3200 *** 148.8600 *** 159.8000 *** 189.5400 *** 204.1400 *** 215.3200 ***
R2 0.9034 0.9014 0.9075 0.9209 0.9261 0.9297
RMSE 473.7200 457.8300 474.8600 469.5600 455.0000 469.0000
DW 2.1427 2.0610 2.1235 2.0799 2.0839 2.0996
AIC 1856.2000 1848.8110 1857.7220 1854.9840 1847.3000 1854.6940
BIC 1875.8280 1871.2430 1880.1540 1877.4170 1869.7320 1877.1260
Results: Dynamic modelDynamic Model by VECM
(1) (2) (3) (4) (5) (6)
y
(1) (2) (3) (4) (5) (6)
constant -21227.5800 -56238.8000 -20634.8000 -21426.1900 -21019.9600 -18632.9300 ***
epi 384.2616 *** 711.6802 390.1137 *** 411.9817 *** 397.8658 *** 357.0456 ***epi 38 6 6 680 390 3 98 39 8658 35 0 56
mpi -28.1819 ** 159.2063 -38.8716 *** -44.9107 *** -39.2171 *** -48.7127 **
im 0.3298 *** 0.4658 0.3603 *** 0.3742 *** 0.3404 *** 0.4935 **
fdi -3.5868 *** -51.2172 *** -2.7098 *** -3.1584 *** -2.9634 *** 2.5925 **
crisis -4787 9550 *** -24365 7800 *** -4494 4680 *** -4814 3490 *** -4595 2110 *** -3989 5580 ***crisis -4787.9550 -24365.7800 -4494.4680 -4814.3490 -4595.2110 -3989.5580
fx 408.2952 *** 7141.1640 *** 315.2183 *** 400.6049 *** 331.9531 *** -2107.9090 ***
reerrisk3 2914.1260 ***
reerrisk6 -111.6965 **
i k9 129 0849 *reerrisk9 -129.0849 *
reerrisk12 -77.5934
reervol -1002.3610 ***
Loglikelihood -3430.122 -3668.363 -3599.785 -3541.434 -3497.759 -3481.497
Overall Test -3430.1220 *** 192.5393 *** 1941.3000 *** 1589.8080 *** 1711.5620 *** 1575.2010 ***
CE Rank 4 5 5 4 4 5
AIC 56.5594 60.5142 59.3899 58.4333 57.7174 54.8766
BIC 57.0191 61.0428 59.9185 58.9620 58.2460 62.9742
Results: Dynamic modelyDynamic Model by VECM
(1) (2) (3) (4) (5) (6)(1) (2) (3) (4) (5) (6)
constant -21227.5800 -56238.8000 -20634.8000 -21426.1900 -21019.9600 -18632.9300 ***
epi 384.2616 *** 711.6802 390.1137 *** 411.9817 *** 397.8658 *** 357.0456 ***epi 38 6 6 680 390 3 98 39 8658 35 0 56
mpi -28.1819 ** 159.2063 -38.8716 *** -44.9107 *** -39.2171 *** -48.7127 **
im 0.3298 *** 0.4658 0.3603 *** 0.3742 *** 0.3404 *** 0.4935 **
fdi -3.5868 *** -51.2172 *** -2.7098 *** -3.1584 *** -2.9634 *** 2.5925 **
crisis -4787 9550 *** -24365 7800 *** -4494 4680 *** -4814 3490 *** -4595 2110 *** -3989 5580 ***crisis -4787.9550 -24365.7800 -4494.4680 -4814.3490 -4595.2110 -3989.5580
fx 408.2952 *** 7141.1640 *** 315.2183 *** 400.6049 *** 331.9531 *** -2107.9090 ***
reerrisk3 2914.1260 ***
reerrisk6 -111.6965 **
i k9 129 0849 *reerrisk9 -129.0849 *
reerrisk12 -77.5934
reervol -1002.3610 ***
Loglikelihood -3430.122 -3668.363 -3599.785 -3541.434 -3497.759 -3481.497
Overall Test -3430.1220 *** 192.5393 *** 1941.3000 *** 1589.8080 *** 1711.5620 *** 1575.2010 ***
CE Rank 4 5 5 4 4 5
AIC 56.5594 60.5142 59.3899 58.4333 57.7174 54.8766
BIC 57.0191 61.0428 59.9185 58.9620 58.2460 62.9742
Conclusions
GARCH is better than Moving variance as a GARCH is better than Moving variance as a volatility measure
Conclusions
GARCH is better than Moving variance as a GARCH is better than Moving variance as a volatility measureDynamic model seems to be better Dynamic model seems to be better (Many variables are non-stationary)
So Thailand’s trade variables might be related in a So Thailand s trade variables might be related in a dynamic way
Conclusions
GARCH is better than Moving variance as a GARCH is better than Moving variance as a volatility measureDynamic model seems to be better Dynamic model seems to be better (Many variables are non-stationary)
So Thailand’s trade variables might be related in a So Thailand s trade variables might be related in a dynamic way
Negative relationship in dynamic model
Conclusions
GARCH is better than Moving variance as a GARCH is better than Moving variance as a volatility measureDynamic model seems to be better Dynamic model seems to be better (Many variables are non-stationary)
So Thailand’s trade variables might be related in a So Thailand s trade variables might be related in a dynamic way
Negative relationship in dynamic model
Remark: We found a positive relationship in static model
Future work
Our planOur planDisaggregated data “Aggregate data might obscure the impact of volatility” gg g g p y
[McKenzie 99]
MVolatility Measure: Use Spot market and Future market exchange ratesBivariate GARCHBivariate GARCH
Future work
Our planOur planDisaggregated data “Aggregate data might obscure the impact of volatility” gg g g p y
[McKenzie 99]
MVolatility Measure: Use Spot market and Future market exchange ratesBivariate GARCHBivariate GARCH
Thank you!