M1and
M2 have been used in this paper to test causality empirically in mean and causality in variance among the
selected stock markets with respect to pre-Brexit and post-Brexit periods.
As
the return series in this paper are found to be heteroskedastic, so unconditional
correlations & partial
correlations are supposed to be biased
upward and do not provide a basis for examining
interdependence. Thus, stock market integration is analysed in the next
part of this paper by utilizing Johansen’s Co-integration method.
4.3 Unit root & Co-integration tests: In this sub section, the pair wise
integrations of the stock markets (to gain
an insight into the existence and extent of co-movement among selected stock
markets) have been tested. Since Brexit have shown its drastic impact on UK stock
market, so to analyse its impact on other stock markets, test of co-integration
is fully justified.
For
existence of co-integration, all the financial time series are supposed to be
integrated of the same order. Hence it
is necessary to confirm the order in which the
index returns are stationary. In
other words, a perquisite of examining co-integration between stock markets is
that all variables are non-stationary. When variable is non-stationary in time
then it is said to have a unit root. Augmented Dickey Fuller (ADF) Test is used
commonly for this purpose. table 7 represents the computations of ADF unit root
test for all the stock indices . Results show that no test statistics are statistically significant
(i.e. the existence of unit roots at level for all daily log return series).
Thus stock indexes (in logarithmic scale)
are integrated. In addition to that,
a further check of differences depicts no evidence to support the
presence of a unit root in first order differences. Thus log returns on indices
are found to be stationary at the first difference. Hence they have been
integrated in the order one.
As the financial series found to be integrated
of the same order, so we can proceed to the next step, which is to test if the
series are co-integrated. Results of bi-variate
Johansen’s Co-integration Test have been presented in table 8 and high degree
of co-integration has been observed.
Table-7:
ADF results
Countries
ADF STATISTIC
At
level
First
Difference
India
-1.800631
-40.90072
Japan
-1.810113
-34.69017
Russia
-2.061185
-43.96004
China
-1.203150
-47.08226
UK
-2.161231
-45.21125
Data
Source: http://?nance.yahoo.com/
Result: Computed using E-Views with respect to the ?rst order differences
in logarithmic stock indices prices.
MacKinnon critical values for rejection of hypothesis of a unit root.:
For the ADF test, at 1% level of
significance Critical Value is -3.4418
; 5% level of significance Critical
Value is -2.8658; 10%
level of significance Critical Value is -2.5691.
Table-8: Bivariate Johansen’s Cointegration Test
Results
Hypothesized Number of Cointegrating Equations
Trace
Statistic
p-value
Max Eigenvalue Statistic
p-value
India-Japan
None
19.897685*
0.0005
17.556685*
0.0075
At most 1
18.807622*
0.0035
16.001181*
0.0045
India-Russia
None
7.007681
0.5015
16.890023*
0.0001
At most 1
17.811185*
0.0002
16.897766*
0.0015
India- China
None
18.898541*
0.0025
17.334285*
0.0031
At most 1
17.000685*
0.0025
16.000483*
0.0022
India-UK
None
6.890185
0.7008
16.004455*
0.0001
At most 1
19.812665*
0.0011
17.891100*
0.0067
Japan-Russia
None
18.891175*
0.0079
4.001156
0.4043
At most 1
17.777685*
0.0035
17.001185*
0.0002
Japan-China
None
8.897319
0.2095
8.000083
0.3022
At most 1
17.899912*
0.0025
17.007685*
0.0003
Japan-UK
None
18.800685*
0.0003
1.111185
0.1015
At most 1
19.555685*
0.0005
18.892431*
0.0025
Russia-China
None
17.811225*
0.0015
16.894445*
0.0002
At most 1
16.135681*
0.0001
16.020185*
0.0003
Russia-UK
None
2.197005
0.4004
1.892017
0.2005
At most 1
19.100085*
0.0045
18.000032*
0.0002
China-UK
None
17.227644*
0.0035
16.123485*
0.0015
At most 1
18.006781*
0.0025
17.000185*
0.0011
Data
Source: http://?nance.yahoo.com/
Result: Computed using E-Views.
* Significant at the 5% level,
**Null Hypothesis (Ho): Series are not cointegrated.
Rejection of null hypothesis implies existence of an underlying relationship of
stock markets.
4.4
Event Study: In this section, using the method of event study, the
impact of Brexit is examined on stock market reaction for the selected
countries. The event of interest for
this paper is Brexit (On 23rd day of June in the year 2016, around 52% of the participating UK electorate
exercised their voting right to leave the EU. On 29th day of March 2017, the
government invoked Article 50 of the Treaty on the
European Union ). Note that
event is not a single date but the period 23rd day of June 2016 to 29th day of
March 2017 has been chosen as event period. The asymmetric event window
has been chosen as -3 (i.e. before) to +5 (i.e. after) days with respect to the
event period.
The
null hypothesis is as follows.
H0
: There is no significant average annual
return (AAR) during the event window caused by happening of Brexit.
To
do event study, Brown & Warner (1980,1985) mentioned three return
generating models like OLS market model or Risk-Adjusted Market
Model (Sharp, 1964) , Market Adjusted Return Model and Mean Adjusted Return Model. This paper uses
only the second one for this purpose.
This model neglects the impact caused by variance in market return in
abnormal return of the security.
Table
reports the results of the event study. It depicts t-statistics of the Market Adjusted Return Model for each day
of the event window. Comparing the
results of the test with the critical values, we conclude that
day +1 , +2 and +3 shows statistical significance; the test
statistics is very high as well as negative and it counts –4.15, -5.12, -5.34, -4.65 and -7.44 for
India, Japan, Russia, China & UK respectively for first day after the event
window. Subsequently these values have been reduced upto +3 day and become
stable there after. It indicates instant significant and negative impact of Brexit
on security prices over all stock markets selected. Note that for any index,
opening prices of all the participating securities have been considered.
Moreover, it is to be noted that the stock market of UK had been badly affected
than other countries, which is quite natural.
Table-9:
Results of Event Study
India
Japan
Russia
China
UK
t-test
(Market Adjusted Return Model)
-3
0.07
0.07
0.02
0.02
0.12
-2
0.07
0.08
0.04
0.09
0.11
-1
0.09
0.01
0.06
0.09
0.09
0
–4.15
-5.12
-5.34
-4.65
-7.44
+1
–4.15
-5.12
-5.34
-4.65
-7.44
+2
–5.17
-6.15
-5.94
-6.05
-7.94
+3
–6.05
-6.92
-6.37
-6.67
-8.37
+4
–0.15
-0.02
-0.04
0.06
0.04
+5
0.01
0.01
0.03
0.06
0.09
Data
Source: http://?nance.yahoo.com/
Result: Computed using Stata.
4.5 Wilcoxon Signed
Ranks Test: Mean return from stock markets before Brexit and after Brexit can be compared using hypothesis testing.
Since the data does not follow normal
distribution (as evidenced from The Jarque–Bera test ), so it is not
recommended to use paired t test. Thus non-parametric equivalent of it, is to
be used. Hence Wilcoxon Signed Ranks Test has been selected for this purpose.
The results are represented in table 10 & table 11.
It
is to be noted that, for all the countries, the p-values are less than 0.05. It
indicates the rejection of the null hypothesis at 5% level of significance.
Hence with 95% confidence, it has been expected that there is a significant
difference in return of stock markets between before Brexit and after Brexit
days. It indicates a serious impact of Brexit on stock markets considered. To
investigate further in detail, econometric models have been considered in later
part of this paper.
Table-10:
Ranks for Wilcoxon Signed Ranks Test
(Before and After Brexit for all countries treated individually)
Ranks
N
India_after_Brexit – India_before_Brexit
Negative Ranks
300a
a. India_after_Brexit < India_before_Brexit Positive Ranks 25b b. India_after_Brexit > India_before_Brexit
Ties
15c
c. India_after_Brexit = India_before_Brexit
Total
340
Japan_after_Brexit – Japan_before_Brexit
Negative Ranks
320d
d. Japan_after_Brexit < Japan_before_Brexit Positive Ranks 20e e. Japan_after_Brexit > Japan_before_Brexit
Ties
0f
f. Japan_after_Brexit = Japan_before_Brexit
Total
340
Russia_after_Brexit – Russia_before_Brexit
Negative Ranks
314g
g. Russia_after_Brexit < Russia_before_Brexit Positive Ranks 20h h. Russia_after_Brexit > Russia_before_Brexit
Ties
6i
i. Russia_after_Brexit = Russia_before_Brexit
Total
340
China_after_Brexit – China_before_Brexit
Negative Ranks
325j
Positive Ranks
15k
j. China_after_Brexit < China_before_Brexit Ties 0l k. China_after_Brexit > China_before_Brexit
Total
340
l. China_after_Brexit = China_before_Brexit
UK_after_Brexit – UK_before_Brexit
Negative Ranks
332m
Positive Ranks
7n
m. UK_after_Brexit < UK_before_Brexit Ties 1o n. UK_after_Brexit > UK_before_Brexit
Total
340
o. UK_after_Brexit = UK_before_Brexit
Data
Source: http://?nance.yahoo.com/
Result: Computed using SPSS and MS Excel with respect to the ?rst order differences
in logarithmic stock indices prices.
