Content uploaded by Erwan Le Saout

Author content

All content in this area was uploaded by Erwan Le Saout on Dec 08, 2019

Content may be subject to copyright.

Bankers, Markets & Investors nº 145 november-december 2016

4

Short-term Impacts of the 2004

Indian Ocean Tsunami on Stock

Markets: A DCC-GARCH Analysis

■I. Introduction

In the past decade, there have been more frequent

occurrences of natural catastrophes that have affected

different parts of the world. Perhaps still fresh in the

mind, is the devastating 2004 Indian Ocean earthquake

and tsunami that killed more than 220,000 people across

all countries affected. Figure 1 shows the main areas affec-

ted by the tsunami. The economic damage resulting from

the tsunami was close to US$9.5 billion in the worst-hit

countries such as Indonesia, Malaysia, Sri Lanka, India

and Thailand (Inderfurth, Fabrycky, and Cohen, 2005).

To take a more recent example, the north-east coast of

Japan was struck by an earthquake and a tsunami in 2011,

which cost at least US$235 billion or around 4% of the

country’s Gross Domestic Product (GDP) according to

World Bank estimates (The Economist, 2011).

The direct and indirect costs of these catastrophic events

are enormous, and better aftermath disaster-management

is needed to curb these potential sums. As a result, there are

emerging interests from industry professionals, as well as

academics who try to address the economic and ﬁ nancial

impacts of natural disasters. On the one hand, academic

studies on the economic impacts of natural disasters are

abundant and widely recognised. For instance, some stu-

dies such as Hallegatte and Dumas (2009), Kim (2011),

Noy (2009), Skidmore and Toya (2002) have examined the

economic consequences of natural disasters in the short

and long runs. On the other hand, to date, studies on the

ﬁ nancial impacts of natural disasters have been limited

and have been primarily focused on speciﬁ c industries or

companies such as the insurance industry and its players.

While there are a few studies on the market-wide effects

of natural disasters, very few have focused only on one

major disaster and its effects on stock markets.

While academic studies have not addressed the in-depth

effects of the 2004 tsunami on stock markets, indepen-

dent organisations have published numerous reports

suggesting that stock markets were not affected during

the post-Indian Ocean tsunami period. In fact, it appears

that stock markets have shown better returns or perfor-

mance. For instance, a report published by the Economist

Intelligence Unit in 2005 showed that stock markets in

affected countries did not collapse. Interestingly, obser-

vations, one month after the disaster, showed that stock

markets in Indonesia, Sri Lanka and Thailand experienced

better performance, compared with the period prior to

the disaster (Economist Intelligence Unit, 2005).

In an attempt to further investigate these phenomena

and to ﬁ ll the gap in the existing literature, this study aims

to examine the short-term effects of natural disasters

on stock markets in the most affected countries – India,

Indonesia, Malaysia and Thailand. The ﬁ ndings of this

study can be used to help professionals to anticipate

* Corresponding author

Adress: 17 rue de la Sorbo nne, 75231 Paris cedex, France

Mail: erwan@l esaout.com

Figure 1. Countries affected by the

2004 Indian Ocean earthquake

ERWAN LE

SAOUT

Maître de

conférences

HDR

Université

Paris 1

Panthéon-

Sorbonne

PRISM

Sorbonne

Labex REFI

SÉBASTIEN

GANNEVAL

Docteur

Université

Paris 1

Panthéon-

Sorbonne

PRISM

Sorbonne

Bankers, Markets & Investors nº 145 november-december 2016 5

SHORT-TERM IMPACTS OF THE 2004 INDIAN OCEAN TSUNAMI ON STOCK MARKETS: A DCC-GARCH ANALYSIS

and manage better in the event of occurrences of natural

disasters. The remainder of the paper proceeds as follows.

Section 1 reviews the existing and related literature on the

topic. Section 2 presents a dynamic-conditional-correla-

tion-multivariate GARCH model. Section 3 shows data

sets and preliminary statistics of stock returns. Section

4 presents our empirical ﬁndings and discussion. The

paper ends with a conclusion.

■II. Literature review

Economic growth has always been a centre of interest

for academics and industry professionals alike. Investors,

in particular, are interested in growing economies and in

identifying opportunities for growth where their wealth

can be grown and expanded.

Natural disasters can be viewed as exogenous or external

shocks that may potentially cause massive destruction to

economies; they are regarded as acting as a kind of Schum-

peterian creative destruction (Cuaresma, Hlouskova, and

Obersteiner, 2008). Some research studies have focused

on their impacts on the short, medium and long-term

growth of economies. Earlier studies of the economic

consequences of natural disasters have concentrated on

short-run responses; for example, Dacy and Kunreuther

(1969) found that GDP tends to increase after a natural

catastrophe occurs. Noy and Vu (2010) examined the

impact of natural disasters on annual output-growth

in Vietnam. More studies in the past two decades have

emerged focusing on short to medium-term impacts of

catastrophes such as Albala-Bertrand (1993a, 1993b),

Noy (2009), and Tol and Leek (1999). From a long-term

perspective, the widely cited work by Skidmore and Toya

(2002) examined long-run relationships between disas-

ters, capital accumulation, total factor productivity and

economic growth. Other studies by Cuaresma, Hlouskova,

and Obersteiner (2008) and Kim (2011) asserted that a

positive correlation between the frequency of natural

disasters and long-run economic growth was found after

conditioning other determinants. However, the ﬁndings

on the long-term effects of disasters are not conclusive.

For instance, an analysis by Hallegatte and Dumas (2009)

found that natural disasters may not necessarily inﬂuence

long-term economic growth.

Other studies aim to assess ﬁnancial impacts of exo-

genous shocks, including natural disasters. Numerous

studies have been done on the impact of exogenous

shocks, from the outbreak of pandemic diseases, to ter-

rorist attacks, and the effect of natural disasters on stock

or equity markets. Intuitively, stock market returns and

volatility differ across countries, and can be affected by

major events around the world. The occurrence of natural

disasters is sudden and unexpected, even though they

can be forecast and anticipated to a degree as a result of

advances in technology. Nevertheless, the consequences

of natural disasters can be intense for the countries that

experience them. It is therefore crucial to possess systematic

studies on the impacts in order that investors and policy

makers alike can better prepare for any repercussions in

ﬁnancial markets after a natural disaster.

At the company level, Lamb (1995, 1998), for example,

investigated the stock price reactions of property and

casualty-related businesses to occurrences of various

hurricanes in the United States. The ﬁndings showed

that the market was able to differentiate between the

magnitudes of hurricanes and the degree of exposure to

loss, which indicates market efﬁciency in incorporating

information. Similarly, Bolak and Suer (2008) observed

negative effects on insurance ﬁrms’ stocks shortly after

an earthquake in Turkey. Shelor, Anderson and Cross

(1990) investigated stock returns for real-estate ﬁrms

after an earthquake in California. They found that returns

are signiﬁcantly negative affected for ﬁrms operating

near to the area affected by the earthquake. Nippani and

Washer (2011) asserted that returns of the top European

airline indices were negatively affected by the eruption of

Eyjafjallajokull, once the effects of global equity markets

and oil prices had been removed.

Beyond analysis at industry and individual company

levels, few studies have also explored the impacts of catas-

trophes on capital markets or, speciﬁcally, equity markets

as a whole. Worthington and Valadkhani (2004) analysed

the impact of natural disasters on the Australian equity

market using daily prices and accumulated returns data

from the All Ordinaries Index. After normalising other

factors, the ﬁndings showed that cyclones, bushﬁres

and earthquakes exerted inﬂuence on market returns,

with cyclones and bushﬁres generally more signiﬁcant.

Worthington (2008), who used a GARCH-M to model

return series, found that natural events and disasters

had no signiﬁcant impact on Australian market returns,

however deﬁned.

The majority of existing research has primarily focused

on analysing a great number of natural disaster occur-

rences and their impacts on capital markets as a whole,

or on specific industries, or even specific companies.

There have been very few studies analysing specific

major disasters and their financial impacts. Nippani

and Bathala (2008) analysed the impact of the Asian

Tsunami on four affected countries. They concluded

that the tsunami had no immediate negative impact

on the Indonesian, Thai, Sri Lankan and Indian stock

markets. The authors used a

t

-test to test changes in

the mean of returns prior to, and after, the tsunami.

Using conditional correlation coefficients, Lee, Wu

and Wang (2007) found no contagion effects between

the stock market indices of countries, but they did find

contagion effects in exchange rate markets, mainly

for a number of Asian countries. Lee, Wu and Wang

(2007) used conditional correlations without a dynamic

model. Moreover, they used a Fisher test to compare

correlations during different periods, while we use

dummy variables.

■III. Methodology

framework

Conditional correlation models are widely employed

in the literature to measure contagion effects and more

speciﬁcally the dynamic conditional correlation model

Bankers, Markets & Investors nº 145 november-december 2016

6

SHORT-TERM IMPACTS OF THE 2004 INDIAN OCEAN TSUNAMI ON STOCK MARKETS: A DCC-GARCH ANALYSIS

(Chiang et al., 2007; Syllignakis and Kouretas , 2011; Min

and Hwang, 2012; Dua and Tujeta, 2016).

Dynamic Conditional Correlation Models, which were

developed by Engle (2002), Engle and Sheppard (2001)

and Tse and Tsui (2002), as an original speciﬁcation for

multivariate models’ conditional correlations, allow the

examination of time-varying dependence between series.

The advantages are numerous because they allow an

estimation of time-varying correlations, which are more

representative of the dynamic than the unconditional

correlation measure or Constant Conditional Correlation

(CCC model) of Bollerslev (1990). Dynamic conditional

correlations rely on two GARCH processes that themselves

capture the time-varying volatilities. The model avoids

computational effort, in contrast with ﬁrst-generation

models, and relies on two parameters (alpha and beta)

to describe the whole dynamic. The DCC-GARCH model

and its extensions deal with ﬁnancial contagion which can

be deﬁned as a signiﬁcant increase in the cross-market

correlation during the crisis (Forbes and Rigobon, 2002).

These models have been used, in particular, to study the

1998 Asian asset-markets crisis (Tse and Tsui, 2002, and

Chiang, Jeon and Li, 2007).

Our approach involves two stages. First we produce an

estimated DCC GARCH model. Second, we model pro-

cesses, following by estimated dynamic correlations. The

estimated DCC-GARCH model is created by a two-step

procedure. In the ﬁrst step, univariate GARCH models

are ﬁtted for each return, and estimates of their variances

are thus obtained. Nevertheless, it should be noted that,

prior to this, autocorrelation of the series is removed using

a VAR (p). In the second step, the standardised residuals

are used to estimate the time-varying correlation matrix.

The multivariate VAR-DCC-GARCH is deﬁned as follows:

X X H

t t p

p

P

t t

= + + ⋅

−

=

∑ε

0

1

1

2

μ (1)

H D R D

t t t t

= (2)

Where

Ht

is the multivariate conditional variance,

X X X X

t t t Nt

= …

( )

1 2

, is the vector of the past observa-

tions, µ0 is a constant, and ε ε ε ε

t t t Nt

= …

( )

1 2

, is the

vector of the standardised residuals.

D diag h h h

t t t NNt

= …

()

11 22

,

is deﬁned as the conditional variance obtained from the

univariate GARCH models.

R diag Q Q Q Q diag Q Q

t t t NNt t t t

= …

()

− − − − −

11

1 2

22

1 2 1 2

11

1 2

22

1 2/ / / / /

, , ……

()

−

QNNt

1 2/

is a time invariant

N N

× symmetric dynamic correla-

tions matrix, where

Q

t is the

N N

× symmetric positive

deﬁnite matrix of standardised residuals εt.

Q a b Q a b Q

t t t t

= − −

( )

+ ⋅ + ⋅

− −

11 1

ε εʹ (3)

Therefore, following Engle (2002), we can write the

correlation coefﬁcients as:

ρij t

ij t

ii t jj t

q

q q

,

,

, ,

=

⋅ where

i j n

, , ,= …1 2 and

i j≠

The speciﬁcation of the model is selected, based on

Log Likelihood, Akaike (AIC) and Schwartz (SIC) infor-

mation criteria.

Then, as Chiang and al. (2007), Min and Hwang (2012)

and Dua and Tujeta (2016) did, we model the estimated

DCC coefﬁcients ρij t, for each of the two indices as an

AR model. This provides further results on the contagion

effect during different periods following the tsunami.

For each correlation series, we estimate three AR models

where dummy variables for different time horizons are

included. These latter are created to represent different

periods after the tsunami, where values of 1 are assigned

for 1 trading day, 5 trading days (one week), and 23 tra-

ding days (one month) after the tsunami; and a value of

0 for all other times.

ρ γ ρ β

ij,t 0

p=1

n

p ij,t-p k k,t

=C + + Dummy +

∑ηij t, (4)

Where

C0

is a constant term,

ρij,t

is the pairwise

conditional correlation between the stock returns of

Asian national indices, and

Dummyk,t

is the dummy

variable, k

†

corresponds to the number of trading days

after the tsunami (

k

=1,

k

=5, and

k

=23 ), and ηij t,.

is a residual term.

■IV. Data and preliminary

analysis

Data sets employed in this study include stock indices’

daily returns for four of the countries most affected by

the tsunami: India, Indonesia, Malaysia and Thailand.

We used extracts of Bloomberg market data from 4

January 2000 to 31 December 2007. This period of study

encompasses the Indian Ocean tsunami that took place

on 26 December 2004.

The Bursa Malaysia KLCI Index was an open capitalisa-

tion-weighted index which comprised some 500 companies

listed on Bursa Malaysia’s Main Board. In 2009, it was

replaced by the FTSE KLCI Index. The Jakarta Stock Price

Index (JCI) is a modiﬁed capitalisation-weighted index

of all stocks listed on the regular board of the Indonesia

Stock Exchange. There were 335 stocks in 2004. The

BSE Sensex Index is a capitalisation-weighted index of 30

well-established, and ﬁnancially sound, companies listed

on the Bombay Stock Exchange (India). The Index was

shifted to a free-ﬂoat methodology on 1 September 2003.

The Bangkok SET 50 Index is a capitalisation-weighted

index of the stocks of the largest 50 companies traded

on the Stock Exchange of Thailand.

Table 1 represents sector allocation for the four Asian

national indices, following the Global Industry Clas-

siﬁcation Standard (GICS). The sectoral distribution

differs depending on the national indices concerned.

Nevertheless, we can observe the signiﬁcant presence of

telecommunications services and the ﬁnancial sector in

at least three indices (JCI, BSE and SET 50).

Bankers, Markets & Investors nº 145 november-december 2016 7

SHORT-TERM IMPACTS OF THE 2004 INDIAN OCEAN TSUNAMI ON STOCK MARKETS: A DCC-GARCH ANALYSIS

Table 3. Testing for

autocorrelation and

heterocedasticity in VAR (3)

residuals

Autocorrelation

Q(5) Prob Q(10) Prob

Portmanteau test 41.13 0.13 117.75 0.34

LM autocorrelation 19.05 0.27 15.35 0.50

Heteroscedasticity

Chi 2 Df Prob

White no cross term 1271.65 240 0

White with cross term 3059.87 900 0

Table 2 reports the descriptive statistics of stock

indices’ daily returns. The Jacques-Bera statistic rejects

the null hypothesis that the returns distribution is

normally distributed. In fact, we can observe nega-

tive skewness in the distribution that indicates that

the observed values have a long tail to the left. Kur-

tosis, for all series, is higher than 3 which indicates

fatter tails in the distribution. These characteristics

Table 1. Sector allocation of Asian stock indices

% KLCI JCI BSE SET 50

Energy 3.43 4.92 15.22 21.78

Materials 5.70 9.46 6.83 23.01

Industrials 21.88 3.51 11.21 8.82

Cons. Discretionary 9.76 10.24 6.55 1.05

Consumer Staples 12.6 16.73 10.31 -

Health Care 0.87 1.98 2.55 -

Financials 8.33 32.31 25.97 25.65

Info. Technology 1.99 0.12 14.01 0.47

Telecom. Services 17.14 19.52 3.91 16.33

Utilities 18.3 1.21 3.44 2.89

Table 2. Summary statistics of the stock index returns

% KLCI JCI BSE SET 50

Mean 0.070% 0.027% 0.066% 0.028%

Standard deviation 1.382% 0.926% 1.540% 1.595%

Skewness -1.00 -0.84 -0.62 -0.55

Kurtosis 10.90 11.29 7.07 12.58

Jarque-Bera 5551.9 5976.9 1513.6 7764.6

suggest that an asymmetric GARCH model should

be more adapted to our series. We also note that BSE

and SET 50 indices are the most volatile, while JCI is

the least so. During the study period, returns of our

four indices are positive: JCI and SET 50 are the less

profitable indices, while the KLCI mean return is the

highest. Augmented Dickey–Fuller tests conclude that

all return series are stationary.

Table 4. Selection of DCC

model

DCC Model Log

Likelihood AIC BIC

GARCH (1,1) 24433.50 -24.38 -24.32

EGARCH (1,1) 24327.60 -24.26 -24.18

GJR (1,1,1) 24484.40 -24.42 -24.36

Bankers, Markets & Investors nº 145 november-december 2016

8

SHORT-TERM IMPACTS OF THE 2004 INDIAN OCEAN TSUNAMI ON STOCK MARKETS: A DCC-GARCH ANALYSIS

■V. Empirical ﬁndings

Prior to the determination of the DCC-GARCH model,

we determined the optimal number of lags of the VAR(p)

model to remove autocorrelation in the return series.

Our results are reported in Table 3. According to the

Portmanteau and LM tests, the null hypothesis of no

serial correlation is not rejected when the number of

selected lags is equal to 3. White tests indicate that the

null hypothesis of no autocorrelation in the squared resi-

duals is rejected. These various results demonstrate that

the presence of volatility clustering and residual series

are heteroscedastic. They justify our choice to retain the

DCC-GARCH model.

Table 4 reports information criteria and likelihood

ratio tests applied on different speciﬁcations of the DCC-

GARCH class – namely, standard GARCH of Bollerslev

(1986), EGARCH of Nelson (1991) and GJR-GARCH of

Glosten et al. (1993). Considering the results, we selected

a DCC model with GJR-GARCH (1,1,1) speciﬁcation for

the equations of conditional variances. Following Table

4, we include a GJR process to account for asymmetrical

behaviour in the correlation process.

h c h I

t t t t t

= + ⋅ + ⋅ + ⋅ ⋅

− − − −

−

0 1 1

2

1 1 1 1

2

1

α ε β γ ε (5)

Where Iif

if

t

t

t

−

−−

−

=<

>

1

1

1

1 0

0 0

,

,

ε

ε

The condition for existence of the second moment

1

of the

GJR-GARCH is observed : this condition is alpha(1) + beta(1)

+ k gamma(1) < 1 (with k = 0.5). This result implies that

the variance is stationary (see Table 5). Volatility is more

persistent for KCLI and BSE (the sum of the coefﬁcients

is closer to 1). The GARCH coefﬁcients are material but

ARCH coefﬁcients are very low. These results show that

recent information has a small inﬂuence in the light of

past information. The GJR coefﬁcients are signiﬁcant for

BSE, JCI and SET50 which means there are asymmetric

effects in volatility. The positive value means that when

the market is bearish, volatility is higher. A Ljung-Box

Q-tests for autocorrelation in the residual series indicate

that there is no heteroscedasticity (Q² (5) and Q² (10)).

The GJR-GARCH estimates have captured the volatility

persistence effect (see Table 5). According to Table 6,

the DCC alpha and beta coefﬁcients are both signiﬁcant

at the 1% level.

Table 7 reports descriptive statistics of dynamic corre-

lations for all pairs. The coefﬁcient is quite low but they

are signiﬁcantly different from zero. The correlations

between the Indonesian, Malaysian and Thailand indices

Table 5. Conditional variance parameter estimates

BSE KCLI JCI SET50

Coef. SE Coef. SE Coef. SE Coef. SE

Constant 0.160*** 0.048 0.482 0.437 0.321*** 0.115 0.363* 0.216

ARCH 0.030 0.020 0.037** 0.015 0.008 0.020 0.039* 0.024

GARCH 0.779*** 0.047 0.938*** 0.028 0.678*** 0.099 0.752*** 0.080

GJR 0.225*** 0.056 0.045 0.030 0.282*** 0.096 0.121** 0.047

Sum 0.921 0.997 0.827 0.852

Q²(5) 4.018 3.311 2.495 0.2

Q²(10) 5.678 4.835 9.464 0.511

Table 6. DCC GJR-GARCH (1,1,1)

model

Coefﬁcient SE

a 0.013*** 0.005

b 0.974*** 0.015

Table 7. Descriptive statistics of dynamic

correlations

BSE-KCLI BSE-JCI BSE-SET50 KCLI-JCI KCLI-SET50 JCI-SET50

Mean 0.175 0.268 0.224 0.293 0.290 0.289

Median 0.174 0.267 0.223 0.289 0.286 0.283

Maximum 0.506 0.534 0.440 0.643 0.534 0.556

Minimum -0.115 0.042 -0.001 0.055 0.071 0.080

Standard

Deviation 0.093 0.086 0.068 0.090 0.073 0.084

Skewness 0.284 0.404 -0.055 0.838 0.232 0.353

Kurtosis 3.805 3.491 3.531 4.381 3.625 3.180

Bankers, Markets & Investors nº 145 november-december 2016 9

SHORT-TERM IMPACTS OF THE 2004 INDIAN OCEAN TSUNAMI ON STOCK MARKETS: A DCC-GARCH ANALYSIS

are the highest. Table 8 reports ADF unit root tests for

conditional correlations. It appears that four series

(KCLI_BSE, JCI_BSE, JCI_KCLI and SET50_JCI) seem

to have a stationary trend. Given these results, we add a

trend in our AR processes to model the dynamic condi-

tional correlations series. Table 9 presents the estimated

coefﬁcients of dynamic correlation processes (Equation

4). Except for the SET50_KCLI conditional correlations

series, the trend is signiﬁcant at 10%. We ﬁnd that the

returns are auto-correlated to order 1. The AR coefﬁcients

are close to 1, as Chiang and al. (2007), Min and Hwang

(2012) and Dua and Tujeta (2016) found. The coefﬁcients

of the dummy variables d1, d5 and d23 are not signiﬁcant

except for the SET50_BSE series. Speciﬁcally, d1 and d5

are signiﬁcant at a 5% level. They indicate a positive effect

of the tsunami on the dependence between Thai index

returns and Indian index returns. Given this set of results,

the contagion effect among Asian national indices in the

post-tsunami period appears to be very low or non-exis-

tent. Figure 2 illustrates this result. Here, the evolution

of dynamic correlations in time and the negligible impact

of the tsunami can be observed.

Our results conﬁrm studies that have concluded that

stock markets in the region generally were not affected

by the disaster and, in fact, behaved differently to what

was expected, compared with performance prior to the

disaster. Such behaviour may seem surprising given that

the media highlighted the unprecedented nature of this

disaster. Nevertheless, it seems that markets behaved

rationally. Tsunami damages had a negligible impact on

Table 8. ADF unit root tests for conditional correlations (DCC-TGARCH)

Constant Trend

Coef. SE Coef. SE Test value

KCLI_BSE 0.001 0.001 1.788E-06** 6.390E-07 -3.967***

JCI_BSE 0.004*** 0.001 3.665E-06*** 7.841E-07*** -5.538***

SET50_BSE 0.003** 0.001 8.417E-07 4.890E-07 -4.150***

JCI_KCLI 0.004** 0.001 1.835E-06** 6.584E-07 -4.352***

SET50_KCLI 0.004** 0.001 6.203E-07 5.276E-07 -4.179***

SET50_JCI 0.004** 0.001 2.139E-06** 6.374E-07 -4.534***

Table 9. Estimation of conditional correlation processes with White standard errors

C Trend AR(1) d1 d5 d23

Coef. SE Coef. SE Coef. SE Coef. SE Coef. SE Coef. SE

KCLI_BSE 0.061 0.049 0.000*** 0.000 0.984*** 0.004 -0.005 0.004

JCI_BSE 0.141*** 0.022 0.000*** 0.000 0.971*** 0.006 -0.001 0.001

SET50_BSE 0.175*** 0.037 0.000* 0.000 0.983*** 0.004 0.002** 0.001

JCI_KCLI 0.191*** 0.042 0.000*** 0.000 0.982*** 0.004 0.002 0.002

SET50_KCLI 0.252*** 0.040 0.000 0.000 0.983*** 0.004 -0.001 0.001

SET50_JCI 0.180*** 0.033 0.000*** 0.000 0.980*** 0.004 0.007 0.006

KCLI_BSE 0.061 0.049 0.000*** 0.000 0.984*** 0.004 0.003 0.009

JCI_BSE 0.141*** 0.022 0.000*** 0.000 0.971*** 0.006 0.001 0.003

SET50_BSE 0.175*** 0.037 0.000* 0.000 0.983*** 0.004 0.002** 0.001

JCI_KCLI 0.191*** 0.042 0.000*** 0.000 0.982*** 0.004 0.000 0.001

SET50_KCLI 0.252*** 0.040 0.000 0.000 0.983*** 0.004 -0.001 0.001

SET50_JCI 0.180*** 0.033 0.000*** 0.000 0.980*** 0.004 -0.000 0.000

KCLI_BSE 0.061 0.049 0.000*** 0.000 0.984*** 0.004 0.007 0.013

JCI_BSE 0.141*** 0.022 0.000*** 0.000 0.971*** 0.006 0.004 0.005

SET50_BSE 0.175*** 0.037 0.000* 0.000 0.983*** 0.004 0.005 0.003

JCI_KCLI 0.191*** 0.042 0.000*** 0.000 0.982*** 0.004 -0.000 0.001

SET50_KCLI 0.252*** 0.040 0.000 0.000 0.983*** 0.004 -0.002 0.001

SET50_JCI 0.180*** 0.033 0.000*** 0.000 0.980*** 0.004 -0.000 0.001

Bankers, Markets & Investors nº 145 november-december 2016

10

SHORT-TERM IMPACTS OF THE 2004 INDIAN OCEAN TSUNAMI ON STOCK MARKETS: A DCC-GARCH ANALYSIS

■VI. Conclusion

This study presents an analysis of the impacts of the

2004 Indian Ocean tsunami on the stock markets of four

affected countries, namely India, Indonesia, Malaysia

and Thailand. We used a dynamic conditional correlation

(DCC) multivariate GARCH model developed by Engle

(2002), Engle and Sheppard (2001) and Tse and Tsui

(2002) in order to capture potential contagion effects

between these markets after the tsunami. The major

advantage of employing this approach is the detection

of potential changes in time-varying conditional cor-

relations. Our results indicate that stock markets in

the region stayed calm, compared with performance

prior to the disaster: natural disasters often provoke

stock market declines, followed by recoveries, when

they occur. This lack of reaction on the Asian financial

markets can be explained by the fact that the damage

was done in outlying areas where there is little concen-

tration of production, and also by the huge amount of

assistance provided which made a positive contribution

to growth. ■

1 An augment ed Dickey–Fuller test (A DF) tests the null hyp othesis of whether a unit

root is present in a conditional v ariance ser ies. Table A1 in the Appendix r eports

ADF tests realized with a const ant and a trend. We reject the unit root hypothe sis at

a 1% level for all series.

the industrial capacity of countries although it caused many

fatalities. The damage was done in outlying areas where

there is little concentration of production. Therefore, there

are few reasons to suppose that companies that make up

the different stock indices would experience declining

results. Furthermore, although it may sound cynical on

the face of it, the disaster may have even produced econo-

mic and commercial beneﬁ ts as a result of the necessary

reconstruction period. Moreover, as the sectoral repre-

sentation is different from one index to another, there

was little probability that the level of correlation increased

signiﬁ cantly following the tsunami. The huge amount of

assistance provided by other countries sensitised by this

tragedy also made a positive contribution to growth. The

tourism and insurance sectors were affected in varying

proportions, but some have claimed that the insurance

industry could ultimately gain from the tsunami due to the

scope for higher premiums and more insurance contracts.

In fact, in Asian countries, companies and individuals

are traditionally underinsured. In addition, the disaster

also boosts the sectors involved in the reconstruction

of coastal areas.

Figure 2. Evolution of dynamic correlations

Bankers, Markets & Investors nº 145 november-december 2016 11

SHORT-TERM IMPACTS OF THE 2004 INDIAN OCEAN TSUNAMI ON STOCK MARKETS: A DCC-GARCH ANALYSIS

■Albala-Bertrand J. (1993a), Natural Disaster

Situations and Growth: A Macroeconomic Model for

Sudden Disaster Impacts, World Development, 21, 1417-

1434.

■Albala-Bertrand J. (1993b), Political Economy of Large

Natural Disasters, Oxford: Clarendon Press.

■Bolak M. and Suer O. (2008), The Effect of Marmara

Earthquake on Financial Institutions, Dogus Universitesi

Dergisi, 9 (2), 135-145.

■Bollerslev T. (1986), Generalized Autoregressive

Conditional Heteroskedasticity, Journal of Econometrics 31,

307-327.

■Box G.E. and Tiao G.C. (1975), Intervention Analysis

with Applications to Economic and Environmental

Problems, Journal of the American Statistical Association, 70

(349), 70-79.

■Chiang T.C. Jeon B. N. and Li H. (2007), Dynamic

Correlation Analysis of Financial Contagion: Evidence

from Asian Markets, Journal of International Money and

Finance, 26 (7), 1206-1228.

■Cuaresma J.C. Hlouskova J. and Obersteiner

M. (2008), Natural Disasters as Creative Destruction?

Evidence from Developing Countries, Economic Inquiry, 46

(2), 214-226.

■Dacy D.C. and Kunreuther H.C. (1969), The Economics

of Natural Disasters, New York: Free Press.

■Dua P. and Tuteja D. (2016), Financial Crises and

Dynamic Linkages across International Stock and

Currency Markets, Economic Modelling, 59, 249-261.

■Economist Intelligence Unit (2005), Af ter the

Tsunami, The Economist.

■Engle R.E. (2002), Dynamic Conditional Correlation: A

Simple Class of Multivariate Generalized Autoregressive

conditional Heteroskedasticity Models, Journal of Business

and Economic Statistics, 20 (2), 339-350.

■Engle R.F. and K. Sheppard (2001), Theoretical

and Empirical Properties of Dynamic Conditional

Correlation Multivariate GARCH, NBER Working Paper

8554.

■Glosten L.R. Jagannathan R. and Runkle D.

(1993), On the Relation between the Expected Value and

The Volatility of the Normal Excess Return on Stocks,

Journal of Finance 48 (5), 1779-1801.

■Hallegat te S. and Dumas P. (2009), Can Natural

Disasters have Positive Consequences? Investigating the

Role of Embodied Technical Change, Ecological Economics,

68, 777-786.

■Inderfurth K.F., Fabrycky D. and Cohen S.P.

(2005), The Tsunami Report Card, December, retrieved

August 5, 2012, from http://www2.gwu.edu/~elliott/

assets/docs/research/reportcard.pdf

■Kim C.K. (2011), The Effects of Natural Disasters on

Long-run Economic Growth, The Michigan Journal of

Business, 4 (1), 11-50.

■Lamb R.P. (1995), An Exposure-based Analysis of

Property-liability Insurer Stock Values around Hurricane

Andrew, The Journal of Risk and Insurance, 62 (1), 111-123.

■Lamb R.P. (1998), An Examination of Market Efﬁciency

around Hurricanes, The Financial Review, 33, 163-172.

■Lee H.-Y., Wu H.-C. and Wang Y.-J. (2007), Contagion

Effect in Financial Markets after the South-East Asia

Tsunami, Research in International Business and Finance, 21

(2), 281-296.

■Min H.-G. and Hwang Y.-S. (2012), Dynamic

Correlation Analysis of US Financial Crisis and

Contagion: Evidence from Four OECD Countries, Applied

Financials Economics, 22 (24), 2063-2074.

■Nelson D.B. (1991), Conditional Heteroskedasticity in

Asset Returns: A New Approach, Econometrica 59 (2), 347-

370.

■Nippani S. and Batha la C.T. (2008), The Impact of

the Asian Tsunami on the Affected Countries’ Stock

Markets, The Ic fai Journal of Applied Finance, 14 (5), 44-

56.

■Nippani S. and Washer K.M. (2011), A note on the

Impact of the Eyjafjallajokull Volcanic Eruption on

Airline Stocks, International Review of Applied Financial

Issues and Economics, 3 (1), 282-295.

■Noy I. (2009), The Macroeconomic Consequences of

Disasters, Journal of Development Economics, 88, 221-231.

■Noy I. and Vu T.B. (2010), The Economics of Natural

Disasters in a Developing Country: The case of Vietnam,

Journal of Asian Economics, 21, 345-354.

■Shelor R.M., Anderson D.C. and Cross M.L. (1990),

The Impact of California Earthquake on Real Estate

Firms’ Stock Value, The Journal of Real Estate Research, 5 (3),

335-340.

■Skidmore M. and Toya H. (2002), Do Natural

Disasters Promote Long-run Growth?, Economic Inquiry,

40 (4), 664-687.

■Syllignakis M.N. and Kouretas G.P. (2011),

Dynamic Correlation Analysis of Financial Contagion:

Evidence from the Central and Eastern European

Markets, International Review of Economics and Finance, 20

(4), 717-732

■The Economist (2011), Natural Disaster: Counting the

Cost, March 21, retrieved from http://www.economist.

com/blogs/dailychart/2011/03/natural_disasters.

■Tol R. and Leek F. (1999), Economic Analysis of

Natural Disasters, In T.E. Downing, A. J. Olsthoorn, and

R. S. Tol, Climate, Change and Risk, London: Routledge, pp.

308-327.

■Tse Y.K. and Tsui A.K.C. (2002), A Multivariate GARCH

Model with Time-varying Correlations, Journal of Business

and Economic Statistics, 20 (2), 351-362.

■Worthington A. C. (2008), The Impact Of Natural

Events And Disasters on the Australian Stock Market:

A GARCH-M Analysis Of Storms, Floods, Cyclones,

Earthquakes and Bushﬁres, Global Business and Economics

Review, 10 (1), 1-10.

■Worthington A.C. and Valadkhani, A. (2004),

Measuring the Impact of Natural Disasters on Capital

Markets: An Empirical Application Using Intervention

Analysis, Applied Economics, 36, 2177-2186.

References

Bankers, Markets & Investors nº 145 november-december 2016

12

SHORT-TERM IMPACTS OF THE 2004 INDIAN OCEAN TSUNAMI ON STOCK MARKETS: A DCC-GARCH ANALYSIS

Table A1. ADF unit root tests for conditional variances (TGARCH)

BSE KCLI JCI SET50

Coef. SE Coef. SE Coef. SE Coef. SE

Constant 3.757e-05 6.389e-06 5.189e-06 1.154e-06 5.918e-05 7.702e-06 5.338e-05 6.586e-06

Trend -9.529e-09 4.791e-09 -2.149e-09 7.8486e-10 -4.491e-09 6.037e-09 -6.775e-09 4.731e-09

ADF -11.262*** -6.359*** -18.097*** -14.408***

The number of lags is determined by the Schwarz criterion

Appendix