IV. DATA AND DATA DESCRIPTION
Tải bản đầy đủ
Glezakos (2007 )). Some previous authors use the daily data in order to update the
dynamic of international transmission (Michael Tucker (1996); Marimuthu (2010)).
Figure 4.1: Movement of daily closing price of stock index from Jan 1st 2005 to Jun
30th 2010
32,000
28,000
24,000
20,000
16,000
12,000
8,000
4,000
0
05
06
07
VNI
S&P
DJI
08
STI
NIK
09
10
SSE
HSI
Examinations of the figure 4.1 reveal that all indices have similar trend: up-trend in
2005-2007 (leading to the positive average daily return); downward trend in 2008first haft of 2009 (leading to the negative average daily return) and up-trend in last
haft of 2009-2010 (leading to the positive average daily return). The closing price of
HSI, NIK, DJI, SSE are far more fluctuant than it of S&P, VNI and STI. HSI is the
most varying while VNI is the least one.
The whole sample period is from January 1st 2005 to June 30th 2010. In order to
examine the effects for the global financial crisis from the year 2007, this sample
data is divided into two sub periods. This first period is pre-crisis (from Jan 1st 2005
to July 22, 2007); the second period is crisis (from July 23, 2007 to Jun 30th, 2010).
The date for the during crisis period is followed the US subprime mortgage crisis
Page 26
started on July 23, 2007 (Adwin (2009); Yellen (2008); Dungey Mardi (2008)).
Whole studied data includes more than 1,300 daily closing price observations of
each stock market. After matching the daily data amongst stock markets for cointegration test performance, the remainders of daily data are about 1,100 daily
observations. It means that over 200 daily observations are taken out of original
data for different none trading days and notional holidays of these stock markets
during the studied period. In addition, time differences are also a vital factor to our
analysis. The trading time of the daily data of price indices is not the same time. For
this reason, the research will be designed that Asian stock price and return at time t
will be run tests with the US’s daily stock price and return at time t-1.
All seven indices are analyzed in daily price and return. The daily returns of each
stock market are gained from logarithmic differences of daily market indices over
the entire sample period and sub periods. The indices are presented in local
currencies. The rationale for this presentation is to evade the problems concerned
with currency transfer since the exchange rate fluctuation in cross countries and
avoid the confounding effect of the regional wide currency devaluation after the
occurrence of the crisis (Yang, (2005)). All estimation results are derived using the
Eviews 6.0 software.
According to Dwyer (1992), the dividends are not dispensable for impacting on the
null hypothesis of the research that no co-integration amongst the mentioned stock
markets. Therefore dividends are not careful consideration for this examination.
Whole data are daily closing price of end of trading day without adjustments.
2. Data descriptive statistics
2.1 Summary statistics
The data descriptive statistic provides the initial information about the nature and
volatility of the stock indices. It also helps observation of how they fare against
Page 27
each other and a comparison of the basis performance indicators of the stock
indices.
A table 4.1 is the summary of basis statistics of the daily stock returns in whole
period. A table 4.2 is summary of basis statistics of the daily stock returns in precrisis period, and table 4.3 is summary of basis statistics of the daily stock returns in
the crisis period respectively. Statistic items shown in these summary statistics are
mean, median, maximum, minimum, standard deviation, skewness, kurtosis and
Jarque-Bera test.
Table 4.1: Summary statistics of the stock returns (Whole period)
VNI
STI
SSE
S&P
NIK
HSI
DJI
Mean
0.000479
-2.18E-05
9.38E-05
-0.000323
-0.000464
0.000295
-0.000292
Median
0.000756
0.000678
0.000549
0.000836
0.000146
0.000799
0.00041
Maximum
0.077505
0.075305
0.090343
0.102457
0.094941
0.134068
0.103259
Minimum
-0.049801
-0.08696
-0.092562
-0.094695
-0.12111
-0.13582
-0.082005
Std. Dev.
0.018292
0.013705
0.019593
0.014574
0.016859
0.018328
0.013479
Skewness
-0.10833
-0.384935
-0.249449
-0.555004
-0.946628
0.049783
-0.298714
Kurtosis
3.530333
8.490096
5.776303
11.83843
10.9322
12.45058
11.99076
Jarque-Bera
14.6867
1375.34
356.0649
3550.908
2976.062
3997.218
3633.28
Probability
0.000647
0
0
0
0
0
0
1074
1074
1074
1074
1074
1074
1074
Observations
It can be clearly seen from the table 4.1 that all stock markets of developed
countries (US and Japan) have negative average daily returns (DJI: -0.0000292;
S&P: -0.000323; NIK: -0.000464) and the stock markets of other countries have
positive average daily returns (VNI: 0.000479; HSI: 0.000295; SSE: 0.0000938)
except for STI (-0.0000218) in the investigation period. In which, VNI has a highest
average daily return followed by HSI, while STI has a lowest one. However the
highest maximum daily return comes from HSI (0.134068) and lowest maximum
daily return comes from STI (0.075305). SSE is the most volatile (0.019593) while
DIJ is the least (0.013479) (as reflected by the standard deviation).
Page 28
A normal distribution is considered to have a zero skewness and kurtosis of three.
The value from above table shows that all daily returns of indices are not normal
distribution. Specifically all daily returns have different zero skewness and almost
negative values except for HSI. The skewness of VNI is less skewed as compared to
that of others. The reason can be explained for that is daily price limit changed from
seven percent to five percent over the period. These negative skewness values
indicate a long left tail. The values of kurtosis are also different with three. Most of
kurtosis values are positive and quite higher than three in this period. Positive and
exceed three values of kurtosis imply a fat tail distribution.
The significant
probability values of Jarque-Bera test once again confirm daily returns have non
normal distribution.
Table 4.2: Summary statistics of the stock returns (Pre-crisis period)
VNI
STI
SSE
Mean
0.001789
0.000772
Median
0.000811
0.001166
S&P
NIK
HSI
DJI
0.001
0.000273
0.000415
0.000695
0.000267
0.000706
0.000864
0.000428
0.001034
0.000415
Maximum
0.077505
0.030573
0.078903
0.021336
0.03522
0.026567
0.020691
Minimum
-0.049801
-0.040367
-0.092562
-0.035343
-0.042304
-0.040793
-0.033488
Std. Dev.
0.016522
0.008434
0.016419
0.00671
0.010527
0.00878
0.006556
Skewness
-0.025825
-0.766934
-0.746109
-0.34611
-0.279356
-0.510736
-0.342649
Kurtosis
4.761022
5.905579
8.048134
4.746291
4.30143
4.715569
4.698732
Jarque-Bera
66.08647
229.8465
590
75.13195
42.70854
84.881
71.44041
0
0
0
0
0
0
0
512
512
512
512
512
512
512
Probability
Observations
It is different with the average daily returns for whole period, all average daily
returns in pre-crisis period are positive. VNI is the most active and profitable
showing the highest average daily return (0.001789). This is followed by SSE
(0.001), STI at 0.000772, HSI at 0.000695, NIK at 0.000415, S&P at 0.000273. The
lowest one comes from DJI (0.000267). Accordingly, the profitable stock markets
are in the developing countries, while those in the developed countries are quite
Page 29
slow. VNI (0.016522) and DIJ (0.006556) continuously keep a most and lowest
volatility. Generally, the stock markets in developing countries are more volatile
than those in the developed countries. Hence, the conventional wisdom in finance of
“the higher risk, higher return” is supported by this results. All stock markets have
negative skewness. Most kurtosis values in pre-crisis period are lower than entire
period. But they are still positive value.
Table 4.3: Summary statistics of the stock returns (Crisis period)
VNI
STI
SSE
S&P
NIK
HSI
-0.000707
-0.000736
-0.000796
-0.000874
-0.001246
-7.43E-05
-0.000812
0.00041
-0.000298
0.000328
0.000738
-0.000233
0
0.000387
Maximum
0.04653
0.075305
0.090343
0.102457
0.094941
0.134068
0.103259
Minimum
-0.048157
-0.08696
-0.074624
-0.094695
-0.12111
-0.13582
-0.082005
Mean
Median
DJI
Std. Dev.
0.01972
0.01713
0.022027
0.019096
0.021009
0.023919
0.017543
Skewness
-0.09833
-0.199479
-0.004742
-0.382377
-0.820863
0.106006
-0.166203
Kurtosis
3.812884
6.311301
4.680412
7.574971
8.303522
8.145164
7.891023
Jarque-Bera
1.725522
260.4843
66.12574
503.8145
721.7628
620.9552
562.7633
Probability
0.421995
0
0
0
0
0
0
562
562
562
562
562
562
562
Observations
All values of mean of average daily return in table 4.3 are consistent with financial
market situation at that time-crisis time. HSI has the lowest return (-0.00000743)
and NIK is the highest one (-0.001246), followed NIK by VNI (-0.000707), STI (.000736), SSE (-0.000796), DJI (-0.000812) and S&P (-0.000874). Most of daily
returns have a negative skewness and positive kurtosis. The HSI (0.023919) and
STI (0.01713) is the most and least volatile.
Page 30
2.2 Correlation
The correlation coefficient is mainly applied to measure whether and how strongly
studied variables are integrated. By using the historical data of daily return,
correlation will help us to examine the short-run relationship amongst indices.
Table 4.4, 4.5, 4.6 list the correlation coefficients amongst seven indices in six stock
markets in whole period, pre crisis and crisis period. DJI and S&P almost have
relatively correlated because they are in same stocks market of US. So we will not
mention about this correlation coefficient in this studying.
Table 4.4: Correlation matrix of average daily return (Whole period)
VNI
VNI
STI
SSE
S&P
NIK
HSI
DJI
STI
SSE
S&P
1 0.11529 0.06361
-0.014
1 0.30506 0.2268
1 0.02992
1
NIK
HSI
DJI
0.21332
0.65383
0.28562
0.10606
1
0.12173
0.77358
0.43779
0.21714
0.68993
1
-0.0171
0.22088
0.02562
0.9848
0.10854
0.21342
1
The evidences from table 4.4 show that there is significantly positive correlation
amongst stock markets. HSI is highly correlated with STI and NIK over the full
period. This implies the group of three stock markets may have a common trend of
same direction. For example, the HSI market go up or go down, the other two
markets will also go up or go down together. Hence, it could be said that portfolio
diversification in three stock markets may not bring benefit to investors. The
correlation of two indices of US stock market show very low correlation with those
Asian stock markets. Especially, the correlation between VNI and S&P, DJI are
negative, which tends to indicate that there is a common trend that is the markets in
the different direction. This is appreciate for international diversification because
Narayan (2005) advices that one condition for international diversification is that
Page 31
the correlation between return should be negative to ensure that some markets will
go up if some go down.
Table 4.5: Correlation matrix of average daily return (Pre-crisis period)
VNI
STI
SSE
S&P
NIK
1 0.07324 -0.0052 -0.0537 0.10538
1 0.20271 0.10351 0.61029
1 0.11093 0.11113
1 0.09272
1
VNI
STI
SSE
S&P
NIK
HSI
DJI
HSI
DJI
0.05572
0.69496
0.25161
0.12596
0.61964
1
-0.0625
0.08832
0.09776
0.96118
0.08444
0.11332
1
In the pre-crisis period, almost average returns are positive and quite low correlation.
Only correlation amongst Vietnam and China and two US indices have negative
correlation. The correlation amongst STI, NIK, HSI still show strongly related in this
time.
Table 4.6: Correlation matrix of average daily return (Crisis period)
VNI
VNI
STI
SSE
S&P
NIK
HSI
DJI
STI
SSE
S&P
NIK
HSI
DJI
1 0.13081 0.09882 -0.0082 0.2586
0.148 -0.0091
1 0.34495 0.24619 0.66243 0.79154 0.2425
1 0.00928 0.35062 0.50653 0.00516
1 0.10737 0.22713 0.98781
1 0.70669 0.11164
1 0.22516
1
All daily returns are positive in the crisis period except for the correlation between
US and Vietnam stock market (VNI and S&P: -0.00818; VNI and DJI: -0.00914) or
Page 32
Vietnam stock market are still different trend with the US stock market in financial
crisis. The others also have correlation increasing. STI, NIK and HSI continuously
have high correlation. We can see positive correlations in table 4.6. Group three
indices of STI, NIK and HSI are always kept a leading position in high correlated
entire period and each sub periods. Conversely, three indices of VNI, S&P and DJI
are low correlation during the studied period. It can be concluded that seven daily
returns are not strong correlation.
Page 33
V. EMPIRICAL RESULTS
1. Analysis of results of unit root test
1.1 The Dickey-Fuller test (DF)
As presented in the previous part, the three stationary tests employed in this
research are DF, ADF and PP test. Since figure 4.1 shows the graphical plot of all
index series trending, the stationary tests use the intercept and trend deterministic
assumption. The appropriate lag length for DF and ADF test is selected using the
Schwarz information criterion (table 5.1). The PP test uses the Bartlett Kernel
estimation method.
Table 5.1: Var lag Order Selection
Lag
0
1
2
3
4
5
6
7
8
LogL
-22438.13
-15795.63
-15644.99
-15583.46
-15547.07
-15508.23
-15473.71
-15437.14
-15397.46
LR
NA
13043.44
291.0113
116.9100
67.97847
71.33146
62.30031
64.82682
69.06672*
FPE
4.80e+35
4.62e+22
2.91e+22
2.75e+22*
2.92e+22
3.06e+22
3.27e+22
3.47e+22
3.63e+22
AIC
102.0233
72.05288
71.59088
71.53392*
71.59125
71.63741
71.70321
71.75972
71.80211
SC
102.0883
72.57302
72.56613**
72.96429
73.47674
73.97802
74.49894
75.01056
75.50807
HQ
102.0490
72.25808
71.97562*
72.09820
72.33507
72.56078
72.80613
73.04218
73.26412
Note: ** Denotes lag order selection by SBIC (at 5% significant level)
The table 5.2 reports the result of DF test (using daily price) of the seven variables
for the entire period of January 2005 to June 2010 and each of the sub periods. The
test statistic is compared with the critical values reported in Fuller (1976) for five
percent significance level. From the results in table 5.2, it can be concluded that the
null hypothesis about the existence of a unit root cannot be rejected for all variables
in the test equation for the entire period as well as for each sub periods. However,
for the first differences of all the variables, the null hypothesis is strongly rejected
for entire and two sub periods. Hence, the conclusion is that all variables are non
stationary in their level forms but stationary in the first differenced forms in entire
Page 34
period and sub periods. The table 5.5 reports the DF test results using the daily
return. We can conclude that all variables are stationary in their level forms in
whole and sub periods.
Table 5.2: DF unit root test (daily price)
Whole period
Variables
Level DF
statistic
First
difference
DF statistic
Pre-crisis period
Level DF
statistic
First
difference
DF statistic
Crisis period
Level DF
statistic
First
difference
DF statistic
VNI
-1.07197
-12.3602
-1.376109
-17.75748
-0.577329
-11.86589
STI
SSE
-0.85311
-0.716824
-35.74019
-13.19513
-1.390935
-0.101434
-23.29285
-23.08226
-0.267005
-1.25167
-15.12077
-24.29501
S&P
NIK
-1.134982
-0.637874
-12.90785
-33.54995
-2.030734
-2.100818
-10.37037
-10.34802
-1.440333
0.186588
-13.34115
-24.54745
HSI
-1.118169
-34.2831
-1.229107
-9.806646
-1.152917
-26.18157
DJI
-1.165513
-14.28462
-1.144802
-10.98457
-1.379545
-14.56015
Note: Denotes significance at the 1%, 5% and 10% level of significance
1.2 The Augmented Dickey-Fuller test (ADF)
Similar to result of DF test, the ADF test is applied to test the seven variables for
unit root. The ADF test results are shown in table 5.3. It could also be concluded
that the test fail to reject the null hypothesis of unit root at level form but we can
reject at order one. This means all time series are not stationary at level form and
integrated of order one in whole and sub periods.
The table 5.6 reports the ADF test results using the daily return. Similarly to the DF
test, we can conclude that all variables are stationary in their level forms in whole
and sub periods.
Page 35
Table 5.3: ADF unit root test (daily price)
Whole period
Variables
VNI
STI
SSE
S&P
NIK
HSI
DJI
Level
ADF
statistic
First
difference
ADF
statistic
1.352507
1.166844
0.674205
1.688941
1.325821
-1.45799
1.560194
Pre-crisis period
Level ADF
statistic
First
difference
ADF
statistic
Crisis period
Level ADF
statistic
First
difference
ADF
statistic
-24.1284
-2.308058
-18.42052
-0.987933
-19.82191
-36.93677
-1.535199
-25.11122
-0.798262
-27.11165
-37.24244
-0.952137
-26.88066
-1.250597
-26.58507
-29.27748
-3.041471
-25.76173
-1.809368
-21.87432
-35.5099
-2.155616
-24.04689
-1.199308
-26.13188
-38.31037
-2.599075
-25.02491
-1.385026
-28.21917
-28.95313
-2.186672
-24.8922
-1.751274
-21.75318
Note: Denotes significance at the 1%, 5% and 10% level of significance
1.3 The Phillips-Perron test (PP)
The results of PP test are displayed in table 5.4 with entire data and sub period’s
data. Seven variables contain a unit root since we can be able to reject the null
hypothesis at 5% level of significance. In summary, the PP unit root tests indicate
the null hypothesis of unit roots is not rejected. Each of daily stock price index data
are integrated of orders one. Thus we can implement the cointegration test using the
daily price to examine the relationship of these indices.
The table 5.7 reports the PP test results using the daily return. Similarly to the DF
and ADF test, we can conclude that all variables are stationary in their level forms
in whole and sub period.
Page 36