International Research Journal of Finance and Economics
ISSN 1450-2887 Issue 98 September, 2012
© EuroJournals Publishing, Inc. 2012
http://www. internationalresearchjournaloffinanceandeconomics.com
Developed and Emerging Stock Markets during the
Global Financial Crisis
Seema Narayan
School of Economics, Finance and Marketing
RMIT University, Melbourne, Australia
E-mail: seema. narayan@rmit. edu. au
Silvia Zia Islam
School of Economics, Finance and Marketing
RMIT University, Melbourne, Australia
E-mail: silvia. islam@rmit. edu. au
Abstract
In this paper, we examine 22 benchmark stock market indices of developed and
emerging countries during the period 31 December 1999-26 February 2010. We study
stock market behaviour focusing on the persistence of market volatility, the leverage effect,
structural breaks or shocks, and market integration and efficiency. The paper finds that
most markets in the two groups behaved similarly in the GFC period. In comparison to the
pre-GFC period, markets became efficient and many experienced a fall in the persistence of
their volatility. The leverage effect, on the other hand, intensified during the GFC. The
dates associated with the structural breaks in stock price indices are found to be same for all
countries on two occasions - February 2007 and August 2008. This is an indication that
market reactions were synchronised. This is unique to the GFC, as our pre-GFC analysis,
which also captures the 2002 market downturns, fails to show similar synchronous
behaviour.
Keywords: Global Financial Crisis (GFC); Stock markets; Efficient capital market
hypothesis; Structural shocks, Leverage effects; Persistence of market
volatility
JEL Classifications Codes: C22, G01, G14, G15
1. Introduction
On the GFC and stock markets, there is currently evidence that 11 major European share markets
experienced significantly higher volatility levels during the GFC covering the period July 2003 to
September 2010 (Milunovich, 2011); for Jordan’s stock returns, volatility is unaffected by the GFC
(2008 and 2009) and the positive relationship between risk and return is preserved in the crisis period
(Al Rjoum, 2011); 8 developed and 2 emerging stock markets are co-integrated during September 2008
to August 2009 (Assidenou, 2011);there is a dramatically stronger correlation between stock markets
of 10 industrialised countries during crisis period from July 2007 to February 2009 and DJ and DAX
were not exerting as much influence on the other 8 indices in the same period (Gklezakoa and
Mylonakis, 2011); for 17 financial markets, volatility plays a key role in explaining the changing
nature of correlations among stock markets over the global financial crisis period from February 2007
International Research Journal of Finance and Economics - Issue 98 (2012)
126
to February 2010 (Mun and Brooks, 2012); and for 8 stock markets of Asia-Pacific, the volatility
spillover effects intensified in the financial crisis period covering the period June 2007 to February
2009 (Chakrabarti, 2011). Further to this, Olowe (2009) documented leverage effect and strong
persistent volatility effect in Nigerian stock markets especially during the GFC period while Mahmood
et al. (2011) indicated that the Chinese stock market was efficient during the GFC.
The focus of our study is on the behaviour of markets during the GFC in terms of, conditional
volatility, market integration, efficiency, and structural breaks in stock prices. Our study is distinct
from the above literature in the following ways. First, to the best of our knowledge, our paper is first to
show synchronization of the major stock markets using structural break analysis and provide insights
about stock market behaviour surrounding the market crash. In particular, our structural break analysis
suggests that that markets had already started to become increasingly synchronised before the stock
market crash in 2008. The structural break dates, derived using the Carrion-i-Silvestre et al. (2005)
test, for all markets converge in February 2007 and/or in August 2008. When these dates are compared
with the stock market indices, we find that by February 2007, markets had not peaked and were still
trending up. Taken together, this is an indication that stock markets began to synchronise well before
markets crashed in 2008. Some recent studies (see above) show market integration from 2007 onwards.
However, it is clear from our results that market reactions were synchronised before and during the
latest stock market crashes. Furthermore, this behaviour is unique to the GFC period when compared
with the 2002 market downturns.
Second, while there are several studies that have gauged market behaviour in the GFC period,
only limited countries have been examined during the GFC period. We examine stock market
behaviour for 22 developing and emerging countries. Furthermore, given that empirical analyses are
often sensitive to the econometric technique or data, it is worthwhile re-examining a bigger set of
countries against different sets of tests.
This paper examines market efficiency using a battery of stationarity tests that test the null of
mean reversion in stock prices against the alternative of a random walk before and during the crisis
period. There is currently on only study on China, that shows mean reversion of stock returns. Our
results indicate that the conventional stationarity tests, Kwaitkowski et al. (KPSS 1992) univariate test
and the Hadri (2000) panel test, provide limited case of mean reversion in stock prices. This result is
consistent with the extant literature, which shows limited case of stationarity. In the traditions of
Perron (1989) to increase power, we use another KPSS type test – the Carrion-i-Silvestre et al. (2005)
test - that allow for endogenous structural breaks in the stock price series to test for mean reversion in
stock prices for all individual countries as well as for two panels – OECD panel and the Asian panel.
The panel tests with structural breaks indicate that market become efficient in the GFC period.
Forth, our paper uses the Exponential Generalised Autoregressive Conditional
Heteroskedasticity (EGARCH) model of Nelson (1991) to examine the volatility of shock returns. This
framework according to Surgailis and Viano (2002:311) ‘… is one of the most successful ARCH
models which may exhibit characteristic asymmetrics of financial time series, as well as long memory.
’ Our results suggest that the volatility of stocks became less persistence for several countries during
the GFC period. The leverage effect is found to be stronger in the GFC period for most stocks.
In what follows, we explain the empirical analysis of this paper. Our empirical results are
discussed in section 4 and some final remarks in the final section.
2. Empirical Analysis
2. 1. Mean Reversion Property and Structural Breaks Tests
It is important to examine market efficiency, as it unravels important information about how the market
behaves. An efficient stock market implies that stock prices are random or that changes in prices are
unpredictable. Mean reverting stock prices have the opposite effect and it implies that short-term
investors would prefer to buy equity when prices are perceived to be lower than their fundamental
127
International Research Journal of Finance and Economics - Issue 98 (2012)
levels. This is also consistent with the contrarian view (DeBondt and Thaler 1989). For long-horizon
investors, this means that the stock market is less risky and that they should invest in equities (Poterba
and Summers 1987). When one examines the empirical literature on the time series properties of stock
market, it becomes clear that evidence in favour of mean reverting stock returns is ample (see, for
instance, Narayan and Smyth 2007; Narayan and Narayan 2007). However, the same is not true for
share prices, which is the focus of the paper here.
Our analysis of the mean reverting property of stock prices includes an array of time series and
panel stationarity tests. The conventional test, the KPSS test, is applied here. However, the short
sample periods, such as ours, does not have enough variability to detect mean reversion. As a result,
panel data models, popularised by the work of Quah (1993) and Levin and Lin (1992) are used to gain
power. In this study we use the Hadri (2000) panel test. Because this test does not guarantee us
individual country results, we apply another test. This is a recent stationarity test with multiple breaks
proposed by Carrion-i-Silvestre et al. (2005). This test follows from the work of Perron (1989), who
showed that the structural break is a source of nonstationarity in a time series process. Carrion-iSilvestre et al. (2005) test is chosen for a number of reasons. First, it is flexible enough to cater for
stationarity test with a levels break (or a break in the intercept) and is, therefore, able to test the null of
mean stationarity. Second, unlike other tests with breaks, this test allows for multiple structural breaks.
Third, it allows for heterogeneity in the panel data setting, and in doing so provides the individual
country result as well as the panel result.
In all, the univariate tests applied in this study includes a stationarity test without breaks,
namely the Kwaitkowski et al. (KPSS, 1992), and a stationarity test with breaks, namely the Carrion-iSilvestre et al. (2005) test. A panel stationarity test without structural break includes the Hadri (2000)
test and a panel test with breaks include Carrion-i-Silvestre et al. (2005) test. 1 We explain each of
these tests below.
The KPSS test, developed by Kwaitkowski et al. (1992), is designed to test the null hypothesis
of stationarity in a series. This test is often seen as a complement to the ADF type tests and is used as a
confirmatory analysis (see Carrion-i-Silvestre, et al. 2005). This technique simply defines the data
generating process of the industrial average stock price series as
SPt t t
(1)
SPt
is described as a sum of a random walk, t , and a stochastic process, t ; where,
t t 1 t ; t is a white noise process 0, 2 , and 0 , is the intercept. Since t and t are
assumed to be mutually independent, the null hypothesis for mean stationarity is tested:
SPt t
(2)
The test statistic is defined as:
ˆ 2 Ŝ2t
i T 2
(3)
t
ˆ
where S t j 1 ˆ j denotes the partial sum process derived from the OLS residuals after estimating
equation (3. 18); ˆ i2 T 1 2t 2 T 1 w s, l ˆ t ˆ t s is the non-parametric estimate of the
T
l
T
t 1
s 1
t s 1
disturbance variance; ws, l is the spectral window; the subscript i denotes any deterministic function
(constant and/or trend). In the case of a test of mean stationarity: f t u with i 1 ; and T is the
sample size.
The KPSS test is constructed with the null hypothesis that SPt is stationary and the alternate
hypothesis that SPt is nonstationary. The null is rejected if the test statistic is greater than the critical
1
The Carrion-i-Silvestre et al. (2005) test performs univariate as well as panel stationarity tests.
International Research Journal of Finance and Economics - Issue 98 (2012)
128
value. In other words, SPt is accepted as a stationary process (in levels or first difference) if the test
statistic is smaller than the critical value.
In a panel setting, the stock price variable is defined as SPi ,t with i 1,....., N individual current
accounts and t 1,....., T . Here, the null hypothesis of stationarity is tested based on:
SPi ,t i i ,t
(4)
Hadri (2000) developed the test of the null hypothesis of stationarity in the panel setting by
averaging the univariate stationarity test of the KPSS test.
If levels shift (or shift in the intercept) is allowed in the above panel setting, the null hypothesis
of stationarity becomes:
SPi ,t i i , k DU i , k ,t i ,t
m
k 1
Here, DU i , k ,t is the dummy variable defined as DU i , k ,t 1 for t T
i
b ,k
(5)
and 0, otherwise; i , t is
k 1,..., m i , m i 1 . As noted above, the model includes individual shifts in the mean caused by
structural breaks. i,k measures the effect of structural breaks on individual series. 2
assumed to be a stationary process; and Tbi,k denotes the kth date of the break for the ith individual,
Carrion-i-Silvestre et al. (2005) extended the Hadri (2000) test of the null hypothesis of
stationarity by accounting for breaks:
LM 1
ˆ 2 T 2 Ŝi2,t
N t
N
T
i 1
Here, Sˆ i ,t j 1 ˆi , j
t
t 1
(6)
, the partial sum process is derived from the OLS residuals after
i ,t
ˆ i2 limT T 1 ES i2,T , i 1,......, N , which allows for heterogeneity across countries. 3 suggests that
estimating
equation
5.
The
long
run
variance
of
calculated
as
above
is,
the test is dependent on the dates of the breaks. For each country, i, it is defined as the vector
i i.1 ,......., i ,m Tbi,1 / T ,....., Tbi,mi / T - this indicates the relative positions of the dates of the
breaks on the entire time period, T.
For the estimation of the break dates, Carrion-i-Silvestre et al. (2005) applied the Bai and
Perron (1998) technique which essentially involves three steps. The first step involves the
identification of the maximum number of break points m max in individual series, which involves
estimating their positions for each m i m max , i 1,...., N . The second step involves testing for the
significance of the breaks for each individual country, i. The third step involves choosing the optimum
number and position for breaks for each country.
As a first step, the Bai and Perron (1998) procedure computes the sum of the squared residuals
(SSR) that is minimized globally. In other words, the arguments chosen for the estimates of the break
dates are the ones that minimize the sequence of individual SSR Tbi,1 ,....,Tbi,m . These are computed
from: Tˆ ,......, Tˆ arg min T ,......,T SSR T ,...., T . Trimming is necessary when computing for
estimates of break dates. The trimming region used here is T [ 0 .1,0 .9 ] . Once all possible dates are
identified, the optimal mi is selected either based on the Bayesian Information Criterion (BIC) and the
i
b ,1
2
3
i
b , mi
i
b ,1
i
b , mi
i
b ,1
i
b,m
Hence, a common restriction on the date of the break, such as Tbi ,k Tb,k , i 1,..., N is not allowed, nor is the restriction
on the number of breaks that an individual time series can have is allowed - m i m j , i j, i , j 1,...., T .
This test can also be performed by assuming homogeneity of the long-run variance across individuals. In this case,
ˆ i2 1 N N i2 .
i 1
129
International Research Journal of Finance and Economics - Issue 98 (2012)
modified Schwarz Information Criterion of Liu, Wu and Zidek (LWZ, 1997) or on the sequential
computation method. Following the Bai and Perron (1998) study, Carrion-i-Silvestre et al. (2005)
recommend LWZ and BIC procedure for trending series and the second approach for non-trending
variables. This second method involves sequential computation and detection of the breaks using the
pseudo F-type test statistic. The number of breaks in this study is chosen using the second procedure.
2. 2. EGARCH Model
Next, we describe the EGARCH model developed by Nelson (1991). 4. We use this framework to
capture the effects of the GFC on the persistence of market volatility and the asymmetric behaviour in
stock prices. 5 A mean equation takes the following form:
ln SPt 1 2 ln SPt 1 t
(7)
Here, 1 and 2 are the parameters to be estimated and t captures disturbances to the
dependent variable that is not observed. The log of the variable, SPt in period t is determined by share
prices of the previous year t 1 .
The variance of the EGARCH model (1, 1)takes the following form:6
t 1
C 5 t 1 C 6 log t21
Log C 3 t 1 C 4
2
t 1
t21
t 1
2
t
(8)
where, C3 is a constant; and C4 - C6 are the other parameters to be estimated. The existence of
asymmetric effects is tested by the hypothesis that C5 0 . If C5 is found to be significantly different
from 0 and C5 0 , then the leverage effect is present. This is when negative shocks make stock prices
more volatile than positive shocks. And in the case, C 0 , positive shocks will make stock prices more
volatile than the negative shocks.
The parameter, C6 , provides information on persistence of stock price volatility. If C6 is close
to one, then, volatility of the share price is implied to be persistent. In other words, shocks to share
prices have long-term effects. If C6 is closer to zero, then, volatility of share prices is persistent and
that the effects of the shocks are felt by short-run share price movements.
5
3. Empirical Results
3. 1. Data
This study employs daily share price indices from 31 December 1999 to 26 February 2010. The time
series data used for the study is sourced from DataStream. The market indices for the respective
countries are as follows: ASX300I, Australia;ATXWBIX, Austria; BGBEL20, Belgium;TTOCOMP,
Canda;
COSEASH,
Denmark;HEXINDX,
Finland;FSBF120,
France;DAXINDX,
Germany;MEREXPI, Italy;TSET100, Japan;KOR200I, South Korea;POPSI20, Portugal;IBEX35I,
Spain; SWEDOMX, Sweden;TRKISTB, Turkey;FTSE100, UK;S&PCOMP, US;CHSCOMP,
China;IBOMBSE, India;FBMKLCI, Malaysia;BNGKS50, Thailand; and TOTXTSG, Singapore. This
data is divided into two sub-samples. Sub-sample 1 covers the pre-GFC period from 4 January 2000 –
4
5
6
Engle (1982) first introduced the volatility theory in the form of the autoregressive conditional heteroskedasticity
(ARCH) model. This was later expanded by Bollerslev (1986), who introduced the generalized ARCH (or GARCH)
framework - see Kim and Wang (2006) for details on the ARCH and GARCH models.
See, Engle and Ng (1993) for discussion on asymmetric behaviourin stock prices and Chou (1988) for persistence of
conditional volatility.
EGARCH (1, 1) is chosen as the preferred model in the interest of parsimony of parameters (see Kim and Wang, 2006).
International Research Journal of Finance and Economics - Issue 98 (2012)
130
30 August 2008, while sub-sample 2 covers the GFC period from 1 September 2008 – 29 January
2010.
3. 2. Descriptive Statistics
We begin our analysis of stock prices in the pre-GFC and the GFC periods using descriptive statistics.
A summary of these statistics for these series as well as the difference between the two periods are
provided in Table 1. The key differences between the GFC period and the pre-GFC period are as
follows. First, the mean values of stock price indices fell for seven countries (the US, UK, Belgium,
Finland, France, Italy and Japan) and increased for rest during the GFC period. Portugal and Sweden
experienced a one per cent increase; Australia, Canada, Denmark, Spain, and Singapore experienced a
20 to 29 per cent increase; Austria and Malaysia experienced a 30 to 40 per cent increase; Thailand
saw a 53 per cent increase; China and Turkey experienced a 83 and 87 per cent increase, respectively;
and India experienced the largest increase at 135 per cent.
Second, the standard deviation of the stock price series, which is a simple measure of volatility
of stocks, did not increase for all the stock market as it is generally assumed during the crisis period.
We find that the standard deviation was higher only for five countries during the GFC period. These
were highest for China (63 per cent) and the US (30 per cent). In terms of the coefficient of variation
(CV), with the exception of Belgium, Italy, Japan, and the US, all the other stock markets experienced
a fall in the CV in the GFC period.
Third, all stocks were positively skewed in the two sample periods but most stock price indices
display a lower coefficient for skewness in the GFC period. The only exceptions to this were France,
Germany and Portugal. Similarly, the kurtosis coefficient for all, except, the UK, Thailand, and France,
fell in the GFC period. This suggests that the tails became thinner during the crisis for most countries.
Table 1:
Descriptive Statistics for the period, 31 December 1999 - 26 February 2010
Table 1 reports the mean, standard deviation, skewness, kurtosis and the coefficient of variation
(CV) of 17 Industrialized and 5 Asian countries during the GFC and pre-GFC period. It also reports
the proportionate change between the two periods.
AUS
Pre-GFC
Mean
Std.Dev.
Skewness
Kurtosis
CV
GFC
Mean
Std.Dev.
Skewness
Kurtosis
CV
GFC/Pre-GFC (% change)
Mean
Std.Dev.
Skewness
Kurtosis
CV
Pre-GFC
Mean
Std.Dev.
Skewness
Kurtosis
CV
JAP
KOR
3889.1 849.3 2974.8 9303.8 274.0 8491.0 3222.8 5155.9 8439.1 974.7
970.4 452.2 747.7 2132.3 85.8 3056.9 743.7 1450.3 1860.7 239.8
1.1
0.9
0.6
0.6
0.9
1.2
0.1
0.1
1.1
0.4
3.1
2.3
2.7
2.3
2.9
3.9
1.8
2.0
3.1
2.2
0.250 0.532 0.251 0.229 0.313 0.360 0.231 0.281 0.220 0.246
118.1
43.7
0.9
2.9
0.370
4807.4 1106.6 2836.9 11679.1 333.2 7669.5 2957.7 5913.6 7434.0 778.9
978.2 386.4 865.8 2039.4 83.9 2398.8 659.4 1156.6 1802.8 210.6
0.3
0.3
0.4
-0.2
0.3
0.6
0.4
0.2
0.9
0.5
2.1
1.6
1.8
1.8
1.9
2.0
2.0
2.0
2.6
1.7
0.203 0.349 0.305 0.175 0.252 0.313 0.223 0.196 0.243 0.270
200.0
33.0
-0.4
2.3
0.165
24
1
-73
-32
-18.5
POR
AUT
30
-15
-67
-30
-34.4
SPA
BEL
-5
16
-33
-33
21.4
SWE
CA
26
-4
-133
-22
-23.8
TUR
DEN
22
-2
-67
-34
-19.6
UK
FIN
-10
-22
-50
-49
-13.1
US
FRA
-8
-11
300
11
-3.4
CHI
GER
15
-20
100
0
-30.5
IND
ITA
-12
-20
69
-3
-12
-24
-18
25
-144
-16
-23
-21
10.0
9.9
-55.4
MAL SINGA THAI
8712.9 9593.1 873.7 21810.2 5275.6 1201.1 1768.9 3297.6 842.7
2307.2 2481.7 263.6 12579.5 869.3 180.3 681.6 1865.1 180.4
0.5
0.6
0.4
0.8
-0.1
-0.1
2.5
1.0
1.1
2.3
2.5
2.2
2.3
1.8
2.3
10.0
2.7
4.2
0.265 0.259 0.302 0.577 0.165 0.150 0.385 0.566 0.214
393.1
124.0
1.4
4.3
0.315
361.9
128.8
0.0
1.4
0.356
131
Table 1:
International Research Journal of Finance and Economics - Issue 98 (2012)
Descriptive Statistics for the period, 31 December 1999 - 26 February 2010 - continued
GFC
Mean
Std.Dev.
Skewness
Kurtosis
CV
GFC/Pre-GFC (% change)
Mean
Std.Dev.
Skewness
Kurtosis
CV
8828.3 11502.7 882.8 40812.1 5184.3 1140.2 3239.6 7759.2 1158.9 476.9
2217.5 2265.2 159.3 10220.5 831.7 233.7 1112.0 1835.6 180.2 113.3
0.6
0.1
0.1
-0.2
0.0
0.1
0.9
-0.4
-0.3
-0.3
2.1
2.1
2.5
1.9
1.9
1.8
2.8
2.3
2.0
1.9
0.251 0.197 0.180 0.250 0.160 0.205 0.343 0.237 0.155 0.238
1
-4
20
-9
-5.1
20
-9
-83
-16
-23.9
1
-40
-75
14
-40.2
87
-19
-125
-17
-56.6
-2
-4
-100
6
-2.6
-5
30
-200
-22
36.5
83
63
-64
-72
-10.9
135
-2
-140
-15
-58.2
38
0
-127
-52
-27.4
552.7
135.7
-0.2
2.1
0.246
21
-9
-121
-56
-24.7
53
5
50
-31.0
3. 3. Mean Reversion
The results on the KPSS (1992) test (without structural breaks) are presented in column 2 of Tables 2,
3 and 4 for the full sample, the Pre-GFC, and the GFC, respectively. These results indicate that stock
prices are generally random walk processes. For the full sample period, the null hypothesis of
stationarity is rejected for all countries at the 1% level of significance, except France, Portugal, the UK
and the US. During the pre-GFC period, the null hypothesis of stationarity is rejected for all the
countries’ stock prices series. And, in the GFC period, all countries stock prices, except Turkey and
India, are found to be non-stationary.
Next, to increase the power of the stationarity test, we group the series into two panels, namely,
the OECD and the Asian panels. The two panels are created for each of the three sample periods
considered in this paper. We then apply Hadri (2000) to test for panel stationarity. The results on the
OECD and Asian panels across the three sample periods are presented in column 2 of Table 5. Our
results show that the null hypothesis of panel stationarity is rejected by the Hadri (2000) test for both
the panels across the full and two sub-sample periods.
The conventional tests generally suggest that most of the stock price indices are random walk.
The absence of evidence of stationarity in time series and panel stock price data is generally consistent
with the literature (see, also for instance, Pindyck 1984; Chou 1988; Narayan and Narayan 2007).
Table 2:
Stationarity test results on stock prices (SP) using univariate KPSS tests with and without breaks
(full sample)
Tables 2, 3 and 4 reports the stationarity test results for the full sample period. These are based on
univariate KPSS test (column 2) and the Carrion-i-Silvestre et al. (2005) test with breaks (column 36). The specification contains country specific intercepts only. The number of break points has been
estimated using the LWZ information criteria, allowing for a maximum of 5 structural breaks. The
bootstrap critical values for this test with breaks are based on 10000 replications and are provided in
the last three columns. The critical values for the KPSS test without breaks at the 5% (1%) level are
0. 463 (0. 739). The KPSS test is constructed with the null hypothesis that SPt is stationary and the
alternate hypothesis that SPt is nonstationarity. The null is rejected if the test statistic is greater than
the critical value. *** and ** denote the level of statistical significance at the 1 % and 5 % level,
respectively.
Country
Australia
Austria
Belgium
KPSS test (without
structural breaks)
Carrion-i-Silvestre test (with
structural breaks)
t-statistic [Bandwidth=41]
t-statistic
4. 08**
4. 271***
0. 979***
0. 012
0. 008
0. 013
Bootstrap Critical Values
(1%)
0. 063
0. 092
0. 061
(5%)
0. 049
0. 068
0. 048
(10%)
0. 043
0. 058
0. 042
International Research Journal of Finance and Economics - Issue 98 (2012)
Table 2:
Stationarity test results on stock prices (SP) using univariate KPSS tests with and without breaks
(full sample) - continued
Canada
Denmark
Finland
France
Germany
Italy
Japan
South Korea
Portugal
Spain
Sweden
Turkey
UK
US
China
India
Malaysia
Singapore
Thailand
Table 3:
132
3. 507***
3. 188***
0. 887***
0. 672
1. 151***
1. 083***
0. 906***
5. 362***
0. 737
2. 361***
0. 854***
5. 310***
0. 677
0. 542
2. 264***
5. 518***
4. 368***
4. 154***
4. 270***
0. 014
0. 009
0. 014
0. 013
0. 013
0. 010
0. 012
0. 015
0. 011
0. 011
0. 015
0. 011
0. 017
0. 012
0. 014
0. 009
0. 015
0. 009
0. 014
0. 092
0. 063
0. 064
0. 070
0. 061
0. 061
0. 062
0. 061
0. 069
0. 087
0. 069
0. 061
0. 062
0. 089
0. 067
0. 060
0. 060
0. 061
0. 061
0. 068
0. 048
0. 049
0. 052
0. 049
0. 048
0. 049
0. 049
0. 052
0. 065
0. 051
0. 048
0. 049
0. 066
0. 051
0. 047
0. 048
0. 049
0. 048
0. 058
0. 042
0. 043
0. 044
0. 042
0. 042
0. 043
0. 042
0. 045
0. 055
0. 044
0. 042
0. 043
0. 057
0. 043
0. 042
0. 042
0. 043
0. 042
Stationarity test results on stock prices (SP) using univariate KPSS tests with and without breaks
(Pre-GFC)
Country
Australia
Austria
Belgium
Canada
Denmark
Finland
France
Germany
Italy
Japan
South Korea
Portugal
Spain
Sweden
Turkey
UK
US
China
India
Malaysia
Singapore
Thailand
KPSS test (without
structural breaks)
Carrion-i-Silvestre test (with
structural breaks)
t-statistic [Bandwidth=36]
t-statistic
4. 601***
5. 378***
2. 666***
3. 240***
3. 734***
1. 473***
1. 348***
1. 314***
3. 854***
1. 313***
4. 532***
2. 252***
1. 337***
4. 844***
1. 367***
1. 387***
4. 601***
0. 792***
4. 575***
3. 442***
4. 138***
4. 980***
0. 015
0. 009
0. 018
0. 018
0. 018
0. 016
0. 025
0. 017
0. 010
0. 019
0. 037
0. 018
0. 010
0. 022
0. 015
0. 022
0. 037
0. 017
0. 020
0. 054
0. 015
0. 011
Critical Values
(1%)
0. 13
0. 133
0. 148
0. 212
0. 148
0. 165
0. 22
0. 156
0. 133
0. 112
0. 162
0. 124
0. 16
0. 219
0. 129
0. 112
0. 217
0. 112
0. 164
0. 189
0. 14
0. 193
(5%)
0. 094
0. 095
0. 101
0. 138
0. 102
0. 11
0. 138
0. 105
0. 095
0. 083
0. 108
0. 091
0. 108
0. 14
0. 094
0. 086
0. 139
0. 082
0. 107
0. 125
0. 097
0. 126
(10%)
0. 077
0. 079
0. 083
0. 107
0. 084
0. 088
0. 107
0. 085
0. 078
0. 071
0. 087
0. 076
0. 087
0. 108
0. 078
0. 072
0. 109
0. 07
0. 086
0. 099
0. 079
0. 099
133
Table 4:
International Research Journal of Finance and Economics - Issue 98 (2012)
Stationarity test results on stock prices (SP) using univariate KPSS tests with and without breaks
(GFC Period)
Country
Australia
Austria
Belgium
Canada
Denmark
Finland
France
Germany
Italy
Japan
South Korea
Portugal
Spain
Sweden
Turkey
UK
US
China
India
Malaysia
Singapore
Thailand
KPSS test (without
structural breaks)
Carrion-i-Silvestre test (with
structural breaks)
t-statistic [Bandwidth =21]
t-statistic
1. 698***
1. 885***
2. 038***
1. 385***
1. 902***
2. 270***
1. 951***
1. 727***
2. 469***
2. 358***
0. 889***
1. 829***
1. 468***
1. 025***
0. 718
1. 521***
1. 789***
1. 319***
0. 738
0. 917***
1. 054***
1. 075***
0. 03
0. 039
0. 03
0. 05
0. 025
0. 018
0. 025
0. 025
0. 026
0. 042
0. 048
0. 032
0. 026
0. 054
0. 019
0. 039
0. 029
0. 045
0. 042
0. 027
0. 055
0. 030
Critical Values
(1%)
0. 192
0. 184
0. 182
0. 188
0. 178
0. 18
0. 185
0. 183
0. 274
0. 186
0. 23
0. 192
0. 182
0. 188
0. 18
0. 194
0. 193
0. 185
0. 178
0. 239
0. 179
0. 183
(5%)
0. 136
0. 131
0. 129
0. 131
0. 126
0. 126
0. 127
0. 129
0. 181
0. 132
0. 153
0. 132
0. 129
0. 131
0. 126
0. 137
0. 132
0. 131
0. 125
0. 158
0. 128
0. 125
(10%)
0. 11
0. 108
0. 105
0. 107
0. 104
0. 104
0. 105
0. 106
0. 14
0. 108
0. 122
0. 108
0. 106
0. 107
0. 104
0. 112
0. 11
0. 106
0. 102
0. 125
0. 104
0. 102
To test whether this non-stationarity is a result of structural breaks, we allow for endogenous
structural breaks in the stock price series. We include five breaks in the full sample, three in the preGFC period and two breaks in the GFC period. The number of breaks selected is dictated by the
sample size. The country specific results on mean stationarity for each of the three sample periods are
presented in column 3 of Tables 2, 3, and 4. The critical values for the full sample, pre GFC and GFC
periods are presented in columns 4-6 of each of the aforementioned tables. Once breaks are introduced
in the stationarity test model, all the series are found to be stationary in all the three samples. This
suggests that markets remained inefficient in the crisis period.
We also provide results on the panel stationarity test with structural breaks. These are presented
in Table 5, columns 3 and 4. We find that markets were inefficient in the period prior to the GFC.
These results are consistent with the univariate results (with breaks) for the full sample and pre-crisis
period. However, for the GFC period, the panel results with structural breaks indicate the OECD and
Asian stock markets were random walks. This test provides some signs that a weak form of market
efficiency may have prevailed during the 2008 financial crisis.
International Research Journal of Finance and Economics - Issue 98 (2012)
Table 5:
134
Results from the panel KPSS type tests with and without breaks (full sample)
The Hadri (2000) panel test results for Pre GFC, and GFC period for the Asian and OECD panels
are presented in column 2 and the Carrion-i-Silvestre et al. (2005) panel test results are presented in
columns 3 and 4 in this table. *** denote statistical significance at 1% level. The probabilities are
reported in parenthesis. The number of break points has been estimated using the LWZ information
criteria, allowing for a maximum of 5 structural breaks for the full sample; 3 structural breaks for
the Pre-GFC period; and 2 structural breaks for the GFC period.
KPSS type test (without breaks)
Carrion-i-Silvestre test (with breaks)
Hadri Z-stat [prob. ]
Homogeneous Bartlett [prob] Heterogeneous Bartlett [prob]
82. 531***
-1. 3360
-1. 0021
OECD (full sample)
[0. 000]
[0. 9092]
[0. 8418]
64. 525***
-0. 500
-0. 350
Asian (full sample)
[0. 000]
[0. 692]
[0. 637]
88. 135***
0. 055
-0. 315
OECD (Pre GFC)
[0. 000]
[0. 478]
[0. 624]
57. 707***
0. 249
0. 916
Asian (Pre GFC)
[0. 000]
[0. 402]
[0. 180]
44. 903***
2. 446***
3. 090***
OECD (GFC)
[0. 000]
[0. 007]
[0. 001]
13. 480***
3. 532***
3. 309***
Asian (GFC)
[0. 000]
[0. 000]
[0. 000]
Panel (sample)
3. 4. Structural Breaks in Stock Prices
In this section, we further examine the structural break dates for the 22 markets. A structural break has
been traditionally seen as a long-term widespread change of the fundamental structure which warrants
a shift in the series, referred to as a structural or regime shift in the series (Hansen 2001). To Hansen
(2001), ‘structural change is a statement about parameters, which only has meaning in the context of a
model’ (p11). As a result, a structural change is captured in a model as a shift in the intercept (known
as a levels shift) or a shift in the trend or both. This view is reflected in the Carrion-i-Silvestre test
which is used in the paper to derive structural breaks in benchmark stock price indices for 22 countries.
Allowing for structural breaks in the study of stock price behaviour is natural given that stock
prices series are, by nature, volatile and sensitive to economic and financial shocks, be it internal or
external. Recent shocks, such as financial crisis, the resulting recession in the US, the economic
slowdown in the US and other Western countries after the September 11 terrorist attacks, and the oil
price hikes may have induced significant levels shifts in the series. Nonetheless, the concept of
structural breaks has been applied, but, only in limited empirical studies relating to empirical asset
pricing and market efficiency (see, Pastor and Stambaugh 2001; Timmermann 2001; Bahng, 2004;
Karoglou, 2010). 7 A recent study by Karoglou (2010) sheds some light on the importance of including
structural breaks in equity markets. This study applies structural breaks in the examination of the
benchmark stock market indices of 27 OECD countries.
The Carrion-i-Silvestre et al. (2005) test allows us to account for a maximum of five breaks in
the intercept for each series. The break dates for the structural shocks are endogenously determined and
are also used in deriving the critical values for the individual and panel tests. The break dates are
presented in Table 6.
These break dates reveal information about market synchronization. First, notice that structural
breaks around the GFC period are same for all the countries. In particular, the February 2007 and
7
Timmermann (2001) provides empirical evidence on the existence of structural breaks in the fundamental process
underlying US stock prices. This paper shows that investor’s knowledge about breaks can strongly affect the dynamics of
asset pricing. Pastor and Stambaugh (2001), on the other hand, estimate the equity premium using the structural breaks,
while Bahng (2004) develops a new mixture-of-normal model in an attempt of applying two-period structural break
analysis.
135
International Research Journal of Finance and Economics - Issue 98 (2012)
August 2008 break dates are same. This means that markets were facing similar shocks or became
synchronized in the period surrounding the crisis.
If these dates are examined against the individual stock price indices, one would notice that
markets were still experiencing an upward trend around the break date in February 2007. Around this
period, there was a striking difference between the developed and emerging markets – while markets
were experiencing increasing returns, the rate of return was already falling in the developed stock
markets but still increasing rapidly in emerging markets. This implies that stock markets had already
begun factoring the future fallout from early 2007. This also implies that markets began to synchronise
well before the markets began to suffer losses.
The market crash from 2008 to early 2009 is captured by the August 2008 break. Recent papers
(highlighted in the introduction) do show increased correlation or co-movement in the GFC period
beginning from 2007. However, our results clearly point out that the integration of the stock market
began before the market crash.
Unlike the structural breaks that relate to the GFC period, the breaks associated with the preGFC period occurred at different times from early 2000s for all the markets examined here. This preGFC period in fact also captures the market downturn of 2002 following the burst of the dot-com
bubble and the 11 September terrorist attacks in the US. These breaks coincided with worldwide
shocks, such as the economic slowdown in the US, the burst of the IT bubble, or oil price shock.
Others have coincided with country-specific shocks relating to the business cycle.
Table 6:
Break dates for the full sample period
Table 6 displays the significant structural break dates derived from the Carrion-i-Silvestre et al.
(2005) test for the 22 stock markets for the full sample. The number of break points has been
estimated using the LWZ information criteria, by allowing for a maximum of 5 structural breaks.
Country
Australia
Austria
Belgium
Canada
Denmark
Finland
France
Germany
Italy
Japan
South Korea
Portugal
Spain
Sweden
Turkey
UK
US
China
India
Malaysia
Singapore
Thailand
No of breaks
5
4
5
4
5
5
5
5
4
5
5
5
4
5
5
5
4
5
5
5
5
5
Break 1
12/11/2001
19/09/2002
25/10/2001
17/09/2002
9/07/2001
19/10/2001
9/7/2001
20/11/2001
22/7/2002
15/8/2001
22/11/2001
19/7/2002
18/7/2002
10/7/2002
25/10/2001
9/10/2001
9/7/2002
27/7/2001
14/11/2001
3/12/2001
3/12/2001
21/12/2001
Break 2
26/05/2003
15/6/2005
5/5/2003
27/4/2004
11/4/2003
29/4/2003
16/1/2003
29/5/2003
21/1/2005
13/6/2003
11/6/2003
27/1/2004
1/7/2004
16/1/2004
11/9/2003
17/4/2003
28/4/2004
2/1/2004
7/8/2003
11/6/2003
11/6/2003
1/7/2003
Break 3
26/11/2004
12/2/2007
15/4/2005
12/2/2007
15/6/2005
4/11/2004
26/7/2004
13/7/2005
12/2/2007
4/8/2005
22/7/2005
4/8/2005
6/2/2007
26/7/2005
4/8/2005
30/6/2005
12/2/2007
4/8/2005
21/7/2005
17/3/2005
28/7/2005
10/3/2005
Break 4
12/2/2007
20/8/2008
12/2/2007
20/8/2008
1/2/2007
12/2/2007
12/2/2007
12/2/2007
20/8/2008
12/2/2007
12/2/2007
12/2/2007
20/8/2008
12/2/2007
12/2/2007
12/2/2007
20/8/2008
12/2/2007
1/2/2007
22/1/2007
5/2/2007
12/2/2007
Break 5
20/8/2008
0
20/8/2008
0
20/8/2008
20/8/2008
20/8/2008
20/8/2008
0
20/8/2008
20/8/2008
20/8/2008
0
20/8/2008
20/8/2008
20/8/2008
0
20/8/2008
20/8/2008
20/8/2008
20/8/2008
20/8/2008
3. 5. Leverage and Persistence
The EGARCH results are summarised in Table 7. Our main results are as follows. We find strong
evidence of leverage effect across all the equity markets studied here. Generally, during bad times one
is inclined to think that negative shocks would lead to more volatility than positive shocks (see, Engle,
1990; Anderson and Bollerslev, 1997; Bollerslev et al. 2006). In this paper, we find evidence in favour
International Research Journal of Finance and Economics - Issue 98 (2012)
136
of the view - that the leverage effect is magnified during bad times - for most of the major stock
markets. With the exceptions of Austria, Sweden, India and Thailand, all the other markets experienced
an increase in the leverage effect during the GFC.
Table 7:
AUS
AUT
BEL
CA
DEN
FIN
FRA
GER
ITA
JAP1
KOR
POR
SPA
SWE
TUR
UK
US
China
India
Malaysia
Singapore
Thailand
E-GARCH framework results
Table 7 displays results derived from the variance of the EGARCH model (1, 1). C(3) is a constant
and C(4)-(6) are the other estimated parameters. Of interest to this paper are C(5) and C(6). The
existence of asymmetric effects is tested by the hypothesis that C 5 0 . The parameter, C 6 ,
provides information on persistence of stock price volatility. The probability associated with each of
the estimated parameters is zero.
C(3)
-0. 297
-0. 530
-0. 358
-0. 181
-0. 293
-0. 104
-0. 238
-0. 244
-0. 771
-0. 303
-0. 252
-0. 456
-0. 252
-0. 187
-0. 265
-0. 219
-0. 188
-0. 194
-0. 564
-0. 320
-0. 304
-0. 932
Full Sample
C(4)
C(5)
0. 144
-0. 114
0. 254
-0. 093
0. 179
-0. 116
0. 103
-0. 070
0. 129
-0. 071
0. 076
-0. 050
0. 103
-0. 123
0. 112
-0. 116
0. 3122
-0. 057
0. 149
-0. 073
0. 135
-0. 075
0. 191
-0. 091
0. 112
-0. 107
0. 089
-0. 096
0. 160
-0. 044
0. 109
-0. 115
0. 083
-0. 114
0. 132
-0. 026
0. 274
-0. 091
0. 167
-0. 066
0. 170
-0. 072
0. 178
-0. 090
C(6)
0. 981
0. 964
0. 976
0. 989
0. 979
0. 994
0. 982
0. 982
0. 945
0. 978
0. 982
0. 967
0. 982
0. 986
0. 981
0. 986
0. 987
0. 988
0. 958
0. 980
0. 981
0. 904
C(3)
-0. 473
-1. 167
-0. 412
-0. 188
-0. 460
-0. 084
-0. 233
-0. 237
-1. 088
-0. 379
-0. 249
-0. 463
-0. 246
-0. 217
-0. 248
-0. 261
-0. 198
-0. 207
-0. 857
-0. 285
-0. 352
-1. 200
Pre-crisis
C(4)
C(5)
0. 120
-0. 136
0. 273
-0. 117
0. 189
-0. 113
0. 086
-0. 063
0. 141
-0. 078
0. 069
-0. 034
0. 088
-0. 121
0. 106
-0. 104
0. 427
-0. 022
0. 153
-0. 073
0. 145
-0. 067
0. 171
-0. 070
0. 103
-0. 093
0. 093
-0. 102
0. 163
-0. 035
0. 099
-0. 131
0. 064
-0. 113
0. 130
-0. 021
0. 296
-0. 114
0. 155
-0. 056
0. 170
-0. 069
0. 176
-0. 087
C(6)
0. 962
0. 901
0. 972
0. 987
0. 963
0. 996
0. 982
0. 983
0. 927
0. 970
0. 983
0. 966
0. 982
0. 983
0. 984
0. 981
0. 984
0. 987
0. 927
0. 982
0. 976
0. 873
C(3)
-0. 484
-0. 340
-0. 435
-0. 220
-0. 180
-0. 158
-0. 306
-0. 247
-0. 247
-0. 213
-0. 212
-0. 644
-0. 312
-0. 100
-0. 443
-0. 247
-0. 207
-0. 572
-0. 560
-0. 402
-0. 235
-0. 418
Crisis
C(4)
C(5)
0. 139
-0. 142
0. 169
-0. 101
0. 159
-0. 136
0. 123
-0. 102
0. 090
-0. 084
0. 086
-0. 086
0. 136
-0. 139
0. 128
-0. 145
0. 2975
-0. 182
0. 086
-0. 104
0. 076
-0. 114
0. 209
-0. 161
0. 128
-0. 149
0. 080
-0. 082
0. 128
-0. 086
0. 089
-0. 122
0. 095
-0. 128
0. 139
-0. 097
0. 246
-0. 089
0. 196
-0. 095
0. 161
-0. 083
0. 173
-0. 062
C(6)
0. 956
0. 974
0. 963
0. 985
0. 987
0. 989
0. 976
0. 982
0. 999
0. 982
0. 981
0. 944
0. 974
0. 995
0. 956
0. 979
0. 984
0. 939
0. 951
0. 973
0. 987
0. 964
The other key result for the GFC is to do with persistence of volatility. Across all three periods,
our results consistently suggest that shocks had a persistent effect on the volatility of major equity
markets of the world. 8 However, for the GFC, we notice a slight change in persistence for almost all
countries. In comparison with the pre-crisis period, we find that volatility was less persistent during the
GFC period for 14 out of 22 countries. These countries are Australia, Belgium, Canada, Finland,
France, Germany, Korea, Portugal, Spain, Turkey, UK, China, Malaysia, and Thailand. For the other 6
out of 22 countries, volatility became more persistent during the GFC period than the pre-GFC period.
This indicates that shocks continued to have a persistent effect on the market in the financial crisis
period.
4. Concluding Remarks
In this study, we examined the behaviour of financial markets in the developed and emerging world in
the period leading to the GFC as well as during the GFC. We found that stock markets in both groups
behaved similarly during the crisis period. We showed that during the GFC, markets in both groups
experienced a fall in persistence of conditional volatility. The leverage effect became more pronounced
during the GFC for markets in the developed and emerging countries. We also found that these markets
became unpredictable or efficient during the GFC.
Our analysis of the structural breaks suggested that structural break dates converged in
February 2007 and August 2008. This synchronization of the markets since February 2007, suggests
8
Poterba and Summers (1986) claimed that shocks to the volatility are transitory and hence, these do not have much
impact on the market. However, later studies (such as, Chou (1988)) that use ARCH-GARCH framework to study the
persistence of stock volatility, generally find that stock volatility is persistent.
137
International Research Journal of Finance and Economics - Issue 98 (2012)
that markets had begun factoring the impending crisis before the fallout in 2008. Market integration
during a financial crisis is commonly found. However, our analysis clearly points out that this
integration between markets around the world began well before the stock market crash. While we did
not find similar patterns during the 2002 market crashes, it may be the case that this is unique to the
GFC period.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
Alexander, C., Lazar, E. (2009) Modelling Regime-Specific Stock Price Volatility, Oxford
Bulletin of Economics and Statistics, 71, 761-97.
Anderson, T. G., Bollerslev, T. (1997) Intraday periodicity and volatility persistence in
financial markets,Journal of Empirical Finance 4, 115-158.
Al Rjoub, S. A. M. (2011) Business Cycles, Financial Crises, and Stock Volatility in Jordon
Stock Exchange, International Journal of Economic Perspective, 5, 83-95.
Assidenou, K. E. (2011) Cointegration of Major StockMarket Indices during the 2008 Global
Financial Distress, International Journal of Economics and Finance 3, 212-222.
Bahng, J. S. (2004) Structural Breaks and the Normality of Stock Returns,Swiss Journal of
Economics and Statistics, 140, 207-227.
Bai, J., Perron P. (1998). Estimating and testing linear models with multiple structural
Change,Econometrica, 66, 47-78.
Balvers, R., Wu, Y., Gilliland, E. (2000) Mean Reversion across National Stock Markets and
Parametric Contrarian Investment Strategies,The Journal of Finance, 55,745-772.
Bollerslev, T., Litvinova, J., Tauchen, G. (2006) Leverage and Volatility Feedback Effects in
High-Frequency Data,Journal of Financial Econometrics,4,353-384.
Carrion-I-Silvestre, J. L., del Barrio, T., López-Bazo, E. (2005) Breaking the Panels: An
Applicationto GDP per capita,Econometrics Journal, 8, 159-175.
Carrion-I-Silvestre, J. L., Sansó A. (2006) A Guide to the Computation of Stationarity Tests,
Empirical Economics,31,433-448.
Chakrabarti, G. ,(2011) Financial crisis and the changing nature of volatility contagion in the
Asia-Pacific region,Journal of Asset Management, 12,172-184.
Chou, Y. R. (1988). Volatility Persistence and Stock Valuations: Some Empirical Evidence
using GARCH, Journal of Applied Econometrics, 3, 279-294.
Debondt, W. F. M., Thaler, R. H. (1989) A Mean-Reverting Walk Down Wall Street, Journal
of Economic Perspective. American Economic Association, 3,189-202.
Engle, R. F. (1982) Autoregressive Conditional Heteroskedasticity With Estimates of the
Variance of United Kingdom Inflation,Econometrica, 50, 987-1008.
Engle, R. F. (1990) Stock Price Volatility and Crash of ’87, Discussion,The Review of
Financial Studies, 3,103-106.
Engle, R. F., Ng, V. K. (1993) Measuring and Testing the Impact of News on Volatility,The
Journal of Finance, XLVIII, 1749-1777.
Fama, E., French, K. (1988). Permanent and temporary components of stock prices, Journal of
Political Economy, 96, 246-273.
Gklezakoa, T., Mylonakis, J. (2011) Links and Interdependence of developed stock markets
under global economic crisis conditions. Journal of Financial Services Marketing, 14, 314-327.
Hadri, K. (2000) Testing for stationarity in heterogeneous panels. The Econometrics Journal,
3,148-161.
Hensen, B. (2001) The New Econometrics for Structural Change: Dating Breaks in U. S. Labor
Productivity,Journal of Economic Perspective,15,117-128.
Jagadeesh, N. (1991) Seasonality in stock price mean reversion: Evidence from the US and the
UK,The Journal of Finance 46, 1427-1444.
International Research Journal of Finance and Economics - Issue 98 (2012)
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
138
Karoglou, M. (2010). Breaking down the non-normality of stock returns,The European Journal
of Finance,16,79-95.
Kwaitkowski, D., Phillips, P. C. B., Schmidt, P., Shin, Y. (1992) Testing the null hypothesis of
stationarity against the alternative of a unit root. How sure are we that economic time series
have a unit root? Journal of Econometrics, 54,159-178.
Levin, A., Lin, C. F. (1992) Unit root in panel data: asymptotic and finite sample Properties.
UCSD, Discussion Paper, 92-123.
Liu, J., Wu, S., Zidek, J. (1997), On Segmented Multivariate Regression,Statistica Sinica 7,
497-525.
Mahmood, F., Xia, X., Ali, M., Usman, M. (2011) How Asian and Global Economic Crisis
Prevailed in Chinese IPO and Stock Market Efficiency,International Business Research, 4,
226-237.
Milunovich, G. (2011) Measuring the impact of the GFC on European equity
markets,Economics Bulletin, 31, 1237-1246.
Mun, M., Brooks, R. (2012) The roles of news and volatility in stock market correlations during
the global financial crisis,Emerging Market Review, 13, 1-7.
Narayan, P. K., Narayan, S. (2007). Mean reversion in stock prices: new evidence from panel
unit root tests,Studies in Economics and Finance, 24, 233-244.
Narayan, P. K., Smyth, R. (2007) Mean reversion versus random walk in G7 stock prices
evidence from multiple trend break unit root tests,Journal of International Financial Markets
Institutions and Money, 17, 152-166.
Nelson, D. B. (1991) Conditional Heteroskedasticity in asset returns: A new approach.
Econometrica, 59, 347-370.
R A. (2009). Stock Return, Volatility And The Global Financial Crisis in an Emerging Market:
The Nigerian Case, International Review of Business Research Papers, 5, 426-447.
Perron, P. (1989) The Great Crash, the oil Shock and the unit root hypothesis,Econometrica 57,
1361-1401.
Pindyck, R. S. (1984) Risk, inflation and the stock market,American Economic Review, 74,
335-351.
Poterba, J., Summers, L. H. (1986) The persistence of volatility and stock market Fluctuations,
American Economic Review, 76, 1142-1151.
Surgaailis, D., Viano, M. C. (2002) Long Memory Properties and Covariance structure of the
EGARCH model,Probability and Statistics, 6, 311-329.
Quah, D. (1993) Empirical cross-section dynamics in economic growth,European Economic
Review,37, 426-434.
Timmermann, A. (2001) Structural Breaks, Incomplete Information and Stock Prices, Journal
of Business and Economic Statistics, 19, 299-314.
View publication stats