REVIEW OF ECONOMIC PERSPECTIVES – NÁRODOHOSPODÁŘSKÝ OBZOR,
VOL. 13, ISSUE 1, 2013, pp. 3–29, DOI: 10.2478/v10135-012-0013-7
To What Extent are Stock Returns Driven by Mean
and Volatility Spillover Effects? – Evidence from Eight
European Stock Markets1
Abdulla Alikhanov2
Abstract: The paper investigates mean and volatility spillover effects from the U.S and
EU stock markets as well as oil price market into national stock markets of eight
European countries. The study finds strong indication of volatility spillover effects from
the US-global, EU-regional, and the world factor oil towards individual stock markets.
While both mean and volatility spillover transmissions from the US are found to be
significant, EU mean spillover effects are negligible. To evaluate the magnitude of
volatility spillovers, the variance ratios are also computed and the results draw to
attention that the individual emerging countries’ stock returns are mostly influenced by
the U.S volatility spillovers rather than EU or oil markets. Additionally, examination of
only global and regional stock markets spillover transmissions into European stock
markets also confirms the dominating presence of the U.S spillover transmissions.
Furthermore, I also implement asymmetric tests on stock returns of eight markets. The
stock market returns of Hungary, Poland, Russia and the Ukraine are found to respond
asymmetrically to negative and positive shocks in the US stock returns. The weak
evidence of asymmetric effects with respect to oil market shocks is found only in the
case of Russia and the quantified variance ratios indicate that presence of oil market
shocks are relatively higher for Russia. Moreover, a model with dummy variable
confirms the effect of European Union enlargement on stock returns only for Romania.
Finally, a conditional model suggests that the spillover effects are partially explained by
instrumental macroeconomic variables, out of which exchange rate fluctuations play the
key role in explaining the spillover parameters rather than total trade to GDP ratios in
most investigated countries.
Key words: Stock markets, the U.S, E.U, volatility spillovers, emerging markets, mean,
oil price, exchange rates, asymmetric effects.
JEL Classification: C26, C51, G15
1. Introduction
Over the last decades, financial markets have experienced dramatic expansion and
interaction with one another. Higher liberalization of economies, globalization and
interrelated synchronization of financial markets have influenced bilateral movements
1
The author would like to thank Frederik Lundtofte and Lu Liu for helpful comments and
suggestions.
2
Lund University, Department of Economics, P.O. Box 7082 S-220 07 Lund, Sweden:
abdulla.alikhan@gmail.com
Volume 13, Issue 1, 2013
4
of equity markets. As the result of globalization and integration and growing
technological advances in financial markets, innovations and shocks in dominant equity
as well as commodity markets are very likely to influence stock returns of emerging
markets. Especially for investors, the behavior and sources of market volatility are of
paramount importance for realization of hedging strategies and international asset
diversification decisions on global financial markets. Additionally, the diversifications
of portfolios of assets are also subject to interlinkages among capital markets. Hence,
the understanding and investigation of this phenomenon is also very crucial for policy
makers.
On the other hand, interrelated development of stock markets across developed and
developing countries has created good opportunities for international investors to invest
in stock markets of emerging economies. Needless to say, the financial markets of the
emerging and developing economies have different characteristics compared to those of
developed countries. For instance, an empirical study by Bekaert and Harvey (1995) on
highly emerging markets which uses data of International Finance Corporation (IFC)
finds out that compared to advanced markets, emerging markets are characterized by
relatively high returns and low correlation. Emerging stock markets seem to be very
appealing for investing since they provide higher expected returns. Besides higher
sample expected returns, distinguishing characteristics of emerging markets are, among
other things, recognized relatively low correlations with mature capital markets and
higher volatility (Harvey 1995). Thus, these differences make a an empirical
investigation of emerging stock markets very appealing, and it isinteresting and valuable
to examine stock returns of European emerging and developing markets within a mean
and volatility spillovers framework.
The purpose of this paper is to examine the mean and volatility spillovers effects from a
global factor US (GF US)3
stock market, regional factor Europe (RF EU) stock market
and as the world factor oil price (WF Oil) changes on the eight European emerging and
developing countries from September 2000 until March 2012. The countries examined
are: Croatia, Czech Republic, Hungary, Poland, Romania, Russia, Turkey, and the
Ukraine. The mean and volatility spillover effects across financial markets are explored
by applying the GJR-GARCH model introduced by Glosten, Jagannathan and Runkle in
1993. Eventually, the calculated variance ratios will allow us to quantitatively analyze
the proportion of volatility spillovers from various sources. Additionally, by excluding
the oil spillover effects, the paper also examines the size and effect of spillover effects
from two markets only: Europe and the US.
Although the spillover models are explored by using the GJR-GARCH model, the
asymmetric tests on stock returns of individual countries will provide for a more
comprehensive investigation of asymmetric existence with respect to any spillover
intensities. Additionally, the paper includes oil price shocks as a world factor to
examine possible spillover effects on stock returns. Moreover, the paper applies
macroeconomic information instruments through conditional spillover model. Finally,
the sensitivity analysis and EU enlargement effect from this study aim to respectively,
3
Throughout the paper, the terms “Global Factor US”, “Regional Factor EU” as well as “World
Factor Oil” are respectively coined with “GF US”, “RF EU” and “WF OIL”
REVIEW OF ECONOMIC PERSPECTIVES
5
contribute an estimation framework and useful information about future expectations for
investors investing other EU candidate stock markets. More specifically, this study aims
to address the following research questions:
1. How do mean and volatility spillover effects of the US, the EU and the oil
market, as a world factor, drive stock returns in European emerging and
developing markets?
2. Which spillover effect has the possibility of having highest magnitude effect on
the selected eight national European stock markets?
3. Does the EU enlargement matter for spillover effects on stock returns?
4. How well are the macroeconomic instruments able to explain global US and
regional EU spillover effects?
2. Literature review
One of the pioneering papers in this field, which investigates volatility spillover effects
of twenty emerging markets in the world is studied by Bekaert and Harvey (1997) who
conclude that global factors drive more volatility effects in the fully integrated markets,
however, in the segmented markets volatility seems to be caused mostly by local
factors. Their study also draws attention to the fact that although volatility appears to be
different in various emerging markets, more liberalized open economies tend to possess
lower volatiles, and capital market liberalization process is the one of the pronounced
driving factor in significant decrease in volatilities.
Furthermore, Rockinger and Urga (2001) explore effects of London and Frankfurt stock
exchange markets on Central European stock markets over the 1994-1997 period. By
applying a similar method proposed by Bekaert and Harvey (1997), they revealed that
although both markets drive significant volatility spillover effects, the effects from UK
stock market tend to be more substantial than those of German stock markets. Another
research by Scheicher (2001) investigates the stock markets of Central and Eastern
European (CEE) countries, namely, Czech, Hungary, and Poland in the light of regional
and global financial market interdependences. Author’s study concludes that equity
markets are influenced by regional and global spillover effects. On the contrary, the
volatility spillovers seem to be driven by regional factors mostly. On the other hand,
Gilmore and McManus (2002) examine short and long run integration and bilateral
relationships between the US and individual CEE stock markets, and find out that
indication of possible interaction is negligible. Applying cointegration tests the
empirical study by Égert and Koubaa (2004) based on GARCH model indicates that
CEE countries are characterized by a higher volatility and more asymmetry than G-7
countries. Moreover, the interactions between three CEE states and developed markets
such as Germany and the US are explored by Syriopoulos (2007). The author finds long
run interactions between developed countries and CEE states. In the short run, however,
US stock market returns impose more dominant effects than those from Germany.
Volume 13, Issue 1, 2013
6
Another research study by Kasman and Torun (2009) investigates the presence of dual
long memory approach proposed by Teyssiere (1997). Their findings show significant
evidence of long memory in time varying variance and mean for CEE stock markets
applying fractionally integrated autoregressive GARCH model. Kocenda and Hanousek
(2010) use highly frequent intraday data to examine spillovers and macroeconomic
news effects from global factor US and regional factor Germany into Czech Republic,
Hungary, and Poland. The authors consider the Frankfurt stock exchange as a regional
factor. They find out that although both of the markets drive strong volatility spillover
transmissions, spillover effects induced by regional Frankfurt Stock is higher than that
of the New York stock exchange market. Along the way, one of the extensive empirical
studies introduced by Beine, Caporale, and Spagnolo (2010) investigates equity markets
of 41 developing countries across the world, and finds out that in most of the countries
equity returns are influenced regional and global spillover effects. Using VAR-MGARCH
(1, 1) model, authors also conclude that Asian and Latin America countries are
more exposed both to return and volatility spillovers, while in emerging European
countries volatility spillovers are main statistical driven stock markets. In addition to
that, if the global spillover effects are dominating in emerging Asian financial markets,
regional spillovers appear to be more pronounced in Latin America and developing
European countries.
Last but not least, Gilmoure et. al., (2006), analyses co-movements of CEE and
developed EU stock market returns. Relying on static and dynamic methods, they report
that co-movements between financial markets have not altered much after EU
membership. On the other hand, behavior of stock returns of “new” EU member
countries was explored by Dvořák and Podpiera (2006), who conclude that some Baltic
and CEE countries’ accession to EU is followed by higher stock market returns. Similar
results are observed in earlier studies for different markets by Henry (2000), Bekaert
and Harvey (2000). Common result is that stock market indexes have been significantly
increased in response to financial market integration.
In summary, we have seen that most empirical studies have focused on developed
markets both across the World and in Europe. After EU enlargement process, some
more new empirical studies have been carried out on Central and Eastern European
countries (CEECs) in recent years. Still, empirical examination of stock markets of
other emerging countries such as Croatia, Russia, the Ukraine and Turkey are underresearched,
which gives cause to the need of further investigation and a deeper analysis
(e.g. Multivariate analysis) which are still being executed.
3. Data description
The data used in this paper were obtained from DataStream International. The raw data
consists of stock indexes of US, aggregate index of EMU countries, crude oil spot prices
and eight stock indexes of eight European countries such as Croatia, Czech Republic,
Hungary, Poland, Romania, Russia, the Ukraine, and Turkey. Sample period of
employed data stock indexes is weekly based and spans from September 1, 2000, to
March, 30, 2012 (Figure 1). In total, the data span includes 604 observations.All indexes
used in the study are in US dollar and were obtained directly from Thomson
International.
REVIEW OF ECONOMIC PERSPECTIVES
7
The broad market index S&P 500 represents the global index, whilst MSCI EU reflects
regional market index, and both of them have been considered to be the so called “broad
based indexes” in US and European developed markets respectively. In addition, MSCI
EU represents all European developed markets.
Figure 1: Indexes and returns
Croatia Czech Republic
Hungary Poland
Romania Russia
-.4
-.3
-.2
-.1
.0
.1
.2
0
200
400
600
800
1,000
1,200
00 01 02 03 04 05 06 07 08 09 10 11 12
CROATIA CROBEX SE - Prices CROATIA CROBEX SE - Returns
-.4
-.3
-.2
-.1
.0
.1
.2
0
20
40
60
80
100
120
00 01 02 03 04 05 06 07 08 09 10 11 12
PRAGUE SE PX - Prices PRAGUE SE PX - Returns
-.4
-.2
.0
.2
.4
0
40
80
120
160
200
00 01 02 03 04 05 06 07 08 09 10 11 12
BUDAPEST (BUX) - Prices BUDAPEST (BUX) - Returns
-.3
-.2
-.1
.0
.1
.2
.3
0
400
800
1,200
1,600
00 01 02 03 04 05 06 07 08 09 10 11 12
W ARSAW SE 20 - Prices W ARSAW SE 20 -Returns
-.4
-.3
-.2
-.1
.0
.1
.2
0
1,000
2,000
3,000
4,000
5,000
00 01 02 03 04 05 06 07 08 09 10 11 12
ROMANIA BET SE - Prices ROMANIA BET SE - Returns
-.4
-.2
.0
.2
.4
0
20
40
60
80
100
00 01 02 03 04 05 06 07 08 09 10 11 12
RUSSIAN MICEX SE- Prices RUSSIAN MICEX SE- Returns
Volume 13, Issue 1, 2013
8
Ukraine Turkey
USA EUROPE
Source: Thomoson Financial Database, 2012
Being captured is 90% of the capitalization of large and liquid securities. According to
Bloomberg, (2012), MSCI index is consists of large and liquid securities, and captures
90% of the capitalization of the broader benchmark. All the indexes are transformed to
returns by taking the difference between log of indexes at time t, and the log of their
own value at time t-1.
The data used for crude oil price is also weekly based. Moreover, macroeconomic
variables employed in this study, such as exchange rate changes of each currency
against Euro and USD, GDP of individual European emerging countries and total trade
between US, EMU countries, and each of the local country is obtained quarterly. The
data for crude prices are extracted from EAI and the macro variables are obtained both
from DataStream and national statistics authorities of each European country.
Descriptive statistics for each of the indexes of different markets are provided in Table
1. All individual countries mean returns are higher than US and European aggregate
returns.
-.4
-.2
.0
.2
.4
.6
0
40
80
120
160
200
240
00 01 02 03 04 05 06 07 08 09 10 11 12
UKRAINE PFTS - Prices UKRAINE PFTS - Returns
-.8
-.4
.0
.4
.8
0
10,000
20,000
30,000
40,000
50,000
60,000
00 01 02 03 04 05 06 07 08 09 10 11 12
ISTANBUL SE 100 - Prices ISTANBUL SE 100 - Returns
-.3
-.2
-.1
.0
.1
.2
600
800
1,000
1,200
1,400
1,600
00 01 02 03 04 05 06 07 08 09 10 11 12
S&P 500 Composite - Prices S&P 500 Composite - Returns
-.3
-.2
-.1
.0
.1
.2
400
800
1,200
1,600
2,000
2,400
00 01 02 03 04 05 06 07 08 09 10 11 12
MSCI EUROPE - Prices MSCI EUROPE - Returns
REVIEW OF ECONOMIC PERSPECTIVES
9
Table 1: Summary statistics of the weekly stock market returns, 09-2000–03-20124
N. of
OBS
Mean Median Max. Min.
Std.
dev.
Skewness Kurtosis
JarqueBera
Test
P-value
Croatia 604 0,192 0,296 16,338 -32,000 3,880 -1,301 13,732 3069,26 0.00
Czech
Rep.
604 0,213 0,614 18,936 -32,780 4,120 -1,244 12,201 2286,56 0.00
Hungary 604 0,218 0,629 20,158 -35,320 4,480 -0,914 9,106 1022,98 0.00
Poland 604 0,219 0,497 24,003 -26,340 4,770 -0,646 6,799 405,47 0.00
Romania 604 0,342 0,474 15,310 -31,590 4,760 -1,224 9,492 1211,87 0.00
Russia 604 0,308 0,759 37,055 -28,720 5,420 -0,457 8,891 892,71 0.00
Ukraine 604 0,298 0,199 41,360 -39,163 7,610 -0,030 9,397 1030,10 0.00
Turkey 604 0,093 0,642 39,680 -75,781 7,470 -1,623 23,072 10404,94 0.00
EMU 604 -0,061 0,282 13,020 -27,340 3,780 -0,986 8,885 969,67 0.00
USA 604 -0,013 11,350 11,350 -20,084 2,700 -0,813 9,737 1208,99 0.00
During the sample period, the highest returns in USD are characterized in Romania and
Russia, respectively, 0.342% and 0.308%. Turkey has a lowest mean return of 0.093%,
followed by Croatia 0.192%. Moreover, the volatility of stock markets is much higher
for the Ukraine, Turkey and Russia with 7.610%, 7.470% and 5.420% standard
deviations, respectively. It should be mentioned that mean returns and standard
deviations for three Central Eastern European (CEE) emerging economies, namely the
Czech Republic, Hungary and Poland are very close to each other. Although the mean
returns of stock markets are higher for developing countries, those markets are
characterized with higher volatilities.
4. Empirical model
The generalized autoregressive conditional heteroskedasticity (GARCH) process is
recognized model for analysis of volatility and return spillovers amongst international
financial markets. The main econometric specification, namely AR(1)-GJR-GARCH
allows to test spillover effects and investigate how much conditional variance individual
country j has been explained respectively by Global Factor US (GF US), Regional
Factor EU (RF EU), local factor (own market of country j) as well as the World Factor
Oil Price (WF Oil).
The core empirical modeling and estimation framework in this paper is based on Ng
(2000), Bekaert et al., (2005), and Christiansen (2004). More specifically, the four
sources, namely pure local shocks of a country j, global US shocks, regional European
4
Note: Weekly returns of indexes are calculated using following formula: ܴ௧ ൌ 100 ∗ ln ቀ
௉೟
௉೟షభ
ቁ ,
where Pt and Pt-1 respectively are logs of indexes at time t, and one period lag.
Volume 13, Issue 1, 2013
10
shocks as well as shocks from oil price innovations are allowed for estimation of
conditional volatility of country j’s stock returns. The models and estimation framework
follow Ng (2000) and Christiansen (2004) approaches mostly. In addition to that, my
paper examines the mean and volatility spillover from oil shocks to country j’s stock
returns, too. Thus, on the four steps univariate autoregressive (AR)GJR-GARCH (1, 1)
is applied. In order to get rid of serial correlation and to avoid ortogonalization, GJRGARCH
model will evolve according to AR (1) process.
5. Constant spillover models
The spillover models are constructed in the following steps indicated below. The first
step is to construct and estimate the bivariate model for Brent oil spot price including a
constant and one period lagged return of oil price.
Step 1: World FactorOil Returns (WF Oil): The unconditional mean for oil returns in
equation (1) changes according to the first order autoregressive (AR) process which
allows us to avoid possible serial correlation. The unconditional variance evolves based
on asymmetric GJR-GARCH (1, 1) which is specified equation (2):
ܴ௢௜௟,࢚ ൌ Ф଴,௢௜௟ ൅ Фଵ,௢௜௟ܴ௢௜௟,௧ିଵ ൅ ߳௢௜௟,௧ (1)
݄௢௜௟,࢚ ൌ ߱௢௜௟ ൅ ܽଵ,௢௜௟ߝ௢௜௟,௧ିଵ
ଶ
൅ ߚ௢௜௟݄௢௜௟,௧ିଵ ൅ ߙଶ,௢௜௟
∗
߳௢௜௟,௧ିଵ
ଶ
‫ܫ‬௢௜௟,௧ିଵ (2)
߳௢௜௟,௧|‫ܫ‬௧ିଵ	~	ܰሺ0, ݄௧ሻ, (3)
The terms Ф଴,௢௜௟and	Фଵ,௢௜௟ are parameter of estimates and ߳௢௜௟,௧ denote a real valued
stochastic process or an idiosyncratic shock which is assumed normally, distributed with
zero mean. The term‫ܫ‬௧ିଵis information set available through time t-1. Subsequently, the
term ݄௢௜௟,௧ stands for conditional variance.
The model enables us to show how volatility behaves differently by having effects of
good and bad news, where:
‫ܫ‬௢௜௟,௧ିଵ ൌ 1 if ߳௢௜௟,௧ ൏ 0,
߱௢௜௟ ൐ 0,
ܽଵ,௢௜௟, ߚ௢௜௟, ܽଵ,௢௜௟ ൅	
ଵ
ଶ
ߙଶ,௢௜௟
∗
൒ 0 and
ܽଵ,௢௜௟ ൅	ߚ௢௜௟ ൅	
ଵ
ଶ
ߙଶ,௢௜௟
∗
൑ 1.
When ߙଶ
∗
positive sign, then “bad” news seems to have a better effect than the “good”
news. In other words, negative shocks have a more noticeable effect on volatility than
positive shocks.
Step 2: Global Factor US (GF US): By following the same previous AR (1)
specification and asymmetric GJR-GARCH model described in Step 1, the mean and
conditional variance for the US stock returns can be through univariate model
respectively, in equations (4) and (5);
ܴ௎ௌ,࢚ ൌ Ф଴,௎ௌ ൅ Фଵ,௎ௌܴ௎ௌ,௧ିଵ ൅ ߮௎ௌܴ௢௜௟,௧ିଵ ൅	ߴ௎ௌ߳௢௜௟,௧ ൅ ߳௎ௌ,௧ (4)
REVIEW OF ECONOMIC PERSPECTIVES
11
݄௎ௌ,࢚ ൌ ߱௎ௌ ൅ ܽଵ,௎ௌ߳௎ௌ,௧ିଵ
ଶ
൅ ߚ௎ௌ݄௎ௌ,௧ିଵ ൅ ߙଶ,௎ௌ
∗
߳௎ௌ,௧ିଵ
ଶ
‫ܫ‬௎ௌ,௧ିଵ (5)
߳௎ௌ,௧|‫ܫ‬௧ିଵ	~	ܰሺ0, ݄௧ሻ, (6)
As in step one, the terms Ф଴,௎ௌand	Фଵ,௎ௌ are parameter of own US stock returns. Step
two also includes effects from one period passed oil price returns, where estimated
parameter is given by in ߮௎ௌ. additionally, and variables ߴ௎ௌ߳௢௜௟,௧ and ߳௎ௌ,௧ respectively,
account for idiosyncratic oil shocks with estimated parameter and own US shocks.
Step 3: Regional Factor EU (RF EU): The univariate model for stock returns of sixteen
European Monetary Union (EMU) countries stock market index can be expressed as
follows:
ܴா௎,࢚ ൌ Ф଴,ா௎ ൅ Фଵ,ா௎ܴா௎,௧ିଵ ൅ ߛ௎ௌܴ௎ௌ,௧ିଵ ൅ ߮ா௎,ܴ௢௜௟,௧ିଵ ൅ ߴா௎߳௢௜௟,௧ ൅ ߠ௎ௌ߳௎ௌ,௧ ൅
߳ா௎,௧ (7)
The equation (7) indicates that European developed markets’ stock returns
correspondingly depend on its own one period lagged returns, one period past US
returns, as well as one period lagged oil returns, where last two lagged returns claim
mean spillovers for European advance stock markets. Eventually, the conditional
variance equation for EU stock returns evolves according to GJR-GARCH (1, 1), where
idiosyncratic EU shock ߳ா௎,௧	 is normally distributed with zero mean;
݄ா௎,࢚ ൌ ߱ா௎ ൅ ܽଵ,ா௎ߝா௎,௧ିଵ
ଶ
൅ ߚா௎݄ா௎,௧ିଵ ൅ ߙଶ,ா௎
∗
߳ா௎,௧ିଵ
ଶ
‫ܫ‬ா௎,௧ିଵ (8)
Eventually, idiosyncratic Oil price shocks and contemporary US residual ߳௎ௌ,௧ account
for volatility spillovers for EU developed market.
Step 4: European Country j’s stock returns: As the final step, the univariate return and
volatility models are specified for each individual European stock returns. As
previously, in order to avoid serial correlation, one period lagged return of each
individual stock return is included and the conditional mean for a country j is thus
given as follows (j=1, 2….8);
ܴ௝,࢚ ൌ ܿ଴,௝ ൅ ܿଵ,௝ܴ௝,௧ିଵ ൅ ߛ௝ܴ௎ௌ,௧ିଵ ൅ ߜ௝ܴா௎,௧ିଵ ൅ ߮௝ܴ௢௜௟,௧ିଵ ൅
൅ߠ௝߳௎ௌ,௧ ൅ ߰௝߳ா௎,௧ ൅	ߴ࢐߳௢௜௟,௧ ൅ ߳௝,௧ (9)
The first row (line) in equation (9) introduces the mean spillovers and parameter
estimates ߛ௝, ߜ௝ and ߮௝ respectively to measure the significance the US, EU and Oil
mean spillovers into each national stock market of European countries. The second raw
(line) in equation (9), presents volatility spillovers from GF US, the RF EU and WF Oil
and the following parameter estimates ߠ௝, ߰௝ and ߴ࢐ empirically test the significance of
volatility spillovers. The idiosyncratic shock, ߳௝,௧ for country j is also normally
distributed and the conditional variance of the residual is specified below, where
݄௝,࢚ ൌ ߱௝ ൅ ܽଵ,௝ߝ௝,௧ିଵ
ଶ
൅ ߚ௝݄௝,௧ିଵ ൅ ߙଶ,௝
∗
߳௝,௧ିଵ
ଶ
‫ܫ‬௝,௧ିଵ (10)
Volume 13, Issue 1, 2013
12
5.1 Volatility spillover effects: Variance ratios
The unexpected returns can also be expressed by ߝ௧ as follows:
ߝ௎ௌ,௧ ൌ	߳௎ௌ,௧ ൅ ߳௢௜௟,௧ (11)
ߝா௎,௧ ൌ	ߠா௎߳௎ௌ,௧ ൅ ߴா௎߳௢௜௟,௧ ൅ ߳ா௎,௧ (12)
Country	jᇱ
s:	ߝ௝,௧ ൌ	ߠ௝߳௎ௌ,௧ ൅ ߰௝߳ா௎,௧߳,௧ ൅ ߴ௝߳௢௜௟,௧߳,௧ ൅ ߳ா௎,௧ (13)
The terms ߳௎ௌ,௧	, ߳ா௎,௧	and	߳௢௜௟,௧ are assumed to be independent by construction. The last
term in equation (13) accounts for remaining effects which are explained by innovations
evolved in individual European stock markets. The conditional variance for unexpected
returns can be calculated by taking variances of different terms in equation (13), and it
is given below:
Country	jᇱ
s:	݄௝.௧ ൌ ‫ܧ‬ሺ߳௎ௌ,௧|߰௧ିଵሻ ൌ ߠ௝
ଶ
݄௎ௌ,௧ ൅ ߰௝
ଶ
݄ா௎,௧ ൅ ߴ௝
ଶ
݄௢௜௟,௧ ൅ ߪ௝,௧
ଶ
(14)
The terms ݄௎ௌ,௧, ݄௢௜௟,௧. ܽ݊݀	݄ா௎,௧ respectively, denote squared US, EU and idiosyncratic
oil market shocks. Consequently, the total variance for individual developing European
Stock market consists of variances of Global Factor US (GF US), Regional Factor EU
(RF EU), and idiosyncratic World Factor Oil price (WF Oil), as well as variance of its
own returns.
urthermore, in order to measure the variances of the unexpected return of country j,
estimated coefficients from unconditional models are utilized. By using equation (13),
the variances can be ultimately computed and the expression for “variance ratios” is
specified as follows:
ܸܴ௝,௧
௎ௌ
ൌ
ఏೕ,೟షభ
మ
ఙೆೄ,೟
మ
௛ೕ.೟
(15)
ܸܴ௝,௧
ா௎
ൌ
ఏೕ,೟షభ
మ
ఙಶೆ,೟
మ
௛ೕ.೟
(16)
ܸܴ௝,௧
ைூ௅
ൌ
ఏೕ,೟షభ
మ
ఙ೚೔೗,೟
మ
௛ೕ.೟
(17)
ܸܴ௝,௧
ை௪௡
ൌ 1 െ ܸܴ௝,௧
௎ௌ
െܸܴ௝,௧
ா௎
െ ܸܴ௝,௧
ைூ௅
(18)
Equations (15), (16), and (17) quantify the variance ratios of various transmitted shocks
in total conditional variances for a country j, and equation (18) simply denotes
remaining variances which are explained by pure local factors.
5.2 Asymmetric spillover tests on stock returns
In order to examine how stock returns of eight (8) individual European countries react
asymmetrically to changes in global and regional stock markets and oil markets, the
equation (19) introduces asymmetric effects of investigated stock markets returns
towards changes from GF US, RF EU and WF Oil effects, in terms of decrease and
increase of stock returns and negative and positive effects of stock innovations on
shocks.
REVIEW OF ECONOMIC PERSPECTIVES
13
ܴ௝,࢚ ൌ ܿ଴,௝ ൅ ܿଵ,௝ܴ௝,௧ିଵ൅	ߛ଴,௝ܴ௎ௌ,௧ିଵ
ି
൅	ߛଵ,௝ܴ௎ௌ,௧ିଵ
ା
൅	ߜ଴,௝ܴா௎,௧ିଵ
ି
൅ ߜଵ,௝ܴா௎,௧ିଵ
ା
൅
߮ଵ,௝ܴ௢௜௟,௧ିଵ
ି
൅ ߮ଵ,௝ܴ௢௜௟,௧ିଵ
ା
൅ ߠ଴,௝߳௎ௌ,௧
ି
൅ ߠଵ,௝߳௎ௌ,௧
ା
൅ ߰଴,௝߳ா௎,௧
ି
൅ ߰ଵ,௝߳ா௎,௧
ା
൅	ߴ଴,௝߳௢௜௟,௧
ି
൅
ߴଵ,௝߳௢௜௟,௧
ା
൅ ߳௝,௧ (19)
In equation (19), both lagged returns and shocks are allowed to have both negative and
positive values. More specifically, the previously modeled mean spillovers are
investigated by decomposing returns	ܴ௧ିଵ in two variables which respectively
represents decrease (ܴ௧ିଵ
ି
ሻ and increase (ܴ௧ିଵ
ା
ሻ in returns of GF US, RF EU and WF Oil
returns. In the same manner, volatility spillovers shocks can be distinguished as ߳௧
ି
and
߳௧
ା
which are a proxy for negative and positive shocks. Subsequently, each of the
variables that accounts for asymmetric effects appeared in equation (19) is constructed
as follows:
ܴ௧ିଵ
ି
ൌ ܴ௧ିଵIfܴ௧ିଵ ൏ 0 and zero (0) otherwise
ܴ௧ିଵ
ା
ൌ ܴ௧ିଵIfܴ௧ିଵ ൐ 0 and zero (0) otherwise
߳௧
ି
ൌ ߳௧If߳௧ ൏ 0 and zero (0) otherwise
߳௧
ା
ൌ ߳௧If߳௧ ൐ 0 and zero (0) otherwise
5.3 Instrumental variables in the spillover model
In addition to unconditional spillover model the conditional spillover driven by
instrumental economic variables introduced by Ng (2000) is also described by equations
(20)-(23). The model includes parameters that relax the time varying constant spillover.
As the degree of international relationships via trades change as well as the degree of
regional integration evolves over time, it is apt to investigate how spillover weight
parameters are captured by some local and regional instrumental variables. For that
reason we allow the following model to account for the notion of conditionality
mentioned above;
ߛ௝,௧ିଵ ൌ ‫ݒ‬௜
ܺ௝,௧ିଵ
௎ௌ
(20)
ߠ௝,௧ିଵ ൌ ‫ݓ‬௜
ܺ௝,௧ିଵ
௎ௌ
(21)
ߜ௝,௧ିଵ ൌ ‫݌‬௜
ܺ௝,௧ିଵ
ா௎
(22)
߰௝,௧ିଵ ൌ ‫ݍ‬௜
ܺ௝,௧ିଵ
ா௎
(23)
Where, ‫ݒ‬ᇱ
and‫ݓ‬ᇱ
are (3x1) vectors of parameters, which quantify the impact of local
information variables on the conditional spillover effects from global factor US, while
‫݌‬ᇱ
and‫ݍ‬ᇱ
are also vector of parameters measuring spillover effects from regional
European effect. The economic variables in ܺ௜,௧ିଵ
௎ௌ
include currency of country j’s
constant exchange rate against US dollar, and total exports from and imports to US as a
ratio of GDP of country j. Subsequently, the variables ܺ௜,௧ିଵ
ா௎
include currency of a
Volume 13, Issue 1, 2013
14
country j’s constant change of exchange rate against EURO, and trade with sixteen
European Monetary Union (EMU 16) countries as ratio of GDP of a country j. It should
be mentioned that all the countries investigated in this paper are either non-Eurozone or
EU members. Currency effects on volatility and correlation of stock markets have been
widely studied. Shocks caused by exchange rate fluctuations can readily transmit to
international financial markets. Moreover, economic integration is defined as the main
motivation behind using the ratios of a trade country j to US and EMU (16) to GDP
(Ng, 2000).
5.4 AR(1)-GJR-GARCH(1,1) – Dummy variable model
The membership in European Union (EU) also led to development of stock exchange
markets of the new member countries. In order to analyse the effect EU Enlargement on
mean and volatility spillovers from regional European market, a dummy variable model
is employed. The dummy variable is created such that, from September 8, 2000, to April
30, 2004, and from September 8, 2000, to December 31, 2006, the dummy 0, and from
May 1, 2000, to March 30, 2012, and from January 8, 2000, to January 1, 2007, the
dummy value 1 is applied respectively to account for the first and second wave of EU
Accession of the aforementioned countries5
. The unconditional AR(1)–GJR–GARCH
(1,1) Dummy variable model and variance quantifications are presented in the equations
below:
ܴ௝,࢚ ൌ ܿ଴,௝ ൅ ܿଵ,௝ܴ௝,௧ିଵ ൅ ߛ௝ܴ௎ௌ,௧ିଵ ൅ ሺ	ߜ଴,௝ ൅ ߜଵ,௝‫ܦ‬௧ିଵሻܴா௎,௧ିଵ ൅ ߮ଵ,௢௜௟ܴ௢௜௟,௧ିଵ ൅
൅ߠ௝߳௎ௌ,௧ ൅ ൫߰଴,௝ ൅ ߰ଵ,௝‫ܦ‬௧ିଵ൯߳ா௎,௧ ൅	ߴ࢐߳௢௜௟,௧ ൅ ߳௝,௧ (24)
ܸܴ௝,௧
ா௎,ௗ௨௠
ൌ
ሺ	టబ,ೕାటభ,ೕ஽೟షభሻ૛ఙ೐ೠ,೟
మ
௛ೕ.೟
(25)
ܸܴ௝,௧
ை௪௡
ൌ 1 െ ܸܴ௝,௧
௎ௌ
െܸܴ௝,௧
ா௎,ௗ௨௠
െ ܸܴ௝,௧
ைூ௅
(26)
Thus, the AR(1)–GJR-GARCH (1,1) dummy model specified in equation (24) and
relevant variance ratios will aim to examine the possible effect (decrease or increase) in
constant spillover parameters.
6. Empirical results
6.1 Constant spillover model
The empirical findings for the mean and return volatility spillovers shown in Table 2
indicate that the first-order autocorrelation parameters of individual countries are
significant for Poland and the Ukraine only. As previously specified in equation (9), the
spillover effects into national stock returns are transmitted by one period lagged returns
and market shocks from the US and EU stock markets, as well as oil market.
5
Czech Republic, Hungary, Poland, Baltic States, Slovakia, and Slovenia joined to EU in the first
wave of EU enlargement in 2004. The second wave defines the membership of Romania and
Bulgaria in 2007.
REVIEW OF ECONOMIC PERSPECTIVES
15
Table 2: The constant spillover model estimation results for an individual country
j’s stock returns over the entire sample period (September 2000 – March 2012)
Croatia Czech6
Hungary Poland Romania Russia Turkey Ukraine
c0 0.288# 0.336* 0.282# 0.207^ 0.359# 0.349# 0.215 0.236
(0.014) (0.001) (0.039) (0.081) (0.018) (0.020) (0.250) (0.267)
[0.116] [0.107] [0.137] [0.119] [0.152] [0.150] [0.187] [0.021]
c1 0.058 -0.070 -0.062 -0.100# -0.025 -0.074 -0.081 0.105#
(0.207) (0.098) (0.148) (0.020) (0.551) (0.061) (0.104) (0.032)
[0.046] [0.042] [0.043] [0.043] [0.044] [0.040] [0.050] [0.049]
ߜ 0.134^ -0.031 0.116 -0.031 0.196# -0.004 0.103 0.464
(0.056) (0.658) (0.141) (0.698) (0.044) (0.958) (0.386) (0.056)
[0.071] [0.070] [0.080] [0.080] [0.092] [0.083] [0.119] [0.078]
ߛ -0.035 -0.166# 0.807 -0.236* -0.121 0.066 -0.002 -0.274#
(0.603) (0.015) (0.067) (0.100) (0.234) (0.528) (0.187) (0.023)
[0.068] [0.068] [0.079] [0.083] [0.101] [0.105] [0.135] [0.121]
߮ 0.024 -0.054# 0.047 0.072* 0.077# 0.180* 0.049 -0.096*
(0.289) (0.020) (0.102) (0.009) (0.015) (0.005) (0.320) (0.007)
[0.023] [0.023] [0.028] [0.028] [0.032] [0.039] [0.050] [0.036]
ψ 0.760* 1.092* 1.239* 1.203* 0.923* 0.701* 1.009* 0.301*
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
[0.066] [0.056] [0.068] [0.071] [0.076] [0.092] [0.120] [0.078]
ߠ 0.431* 0.780* 0.866* 0.979* 0.605* 0.782* 1.046* 0.280*
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
[0.045] [0.035] [0.039] [0.041] [0.058] [0.051] [0.088] [0.063]
ߴ 0.138* 0.179* 0.202* 0.231* 0.227* 0.364* 0.212* 0.069
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.175)
[0.025] [0.023] [0.031] [0.027] [0.049] [0.037] [0.050] [0.051]
ω 1.687 0.379 2.589 0.341# 0.389^ 0.499 0.189# 0.947
(0.000) (0.005) (0.001) (0.039) (0.052) (0.006) (0.050) (0.002)
[0.466] [0.133] [0.080] [0.165] [0.021] [0.182] [0.096] [0.303]
α 0.158* 0.125* 0.063 0.115* 0.104* 0.122* 0.029# 0.236*
(0.002) (0.000) (0.215) (0.003) (0.000) (0.002) (0.038) (0.000)
[0.052] [0.037] [0.051] [0.039] [0.031] [0.040] [0.014] [0.039]
α* 0.108 -0.027 0.161# -0.005 -0.020 0.001 0.058* 0.06
(0.130) (0.468) (0.025) (0.893) (0.454) (0.998) (0.001) (0.287)
[0.071] [0.038] [0.071] [0.041] [0.027] [0.037] [0.018] [0.061]
β 0.606* 0.836* 0.619* 0.853* 0.884* 0.849* 0.936* 0.767*
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
[0.072] [0.039] [0.098] [0.036] [0.028] [0.027] [0.007] [0.017]
Note: P values and std.errors are in the parentheses and brackets, whilst *, #, and ^ represent
significance level at 1%, 5% and 10%, respectively
6
The Czech Republic appears in Tables abbreviated as Czech only.
Volume 13, Issue 1, 2013
16
According to the scale and p-values of the parameter estimated, and the strong
evidences of volatility spillovers from the US and EU stock markets, as well as oil
market, are found to be highly significant for all national stock markets (p-value of
0.000) The parametric Wald tests of no volatility spillovers on stock returns are also
rejected (Table 3). The null hypotheses of joined Wald tests employed for the purpose
of estimation is given below;
H଴
ଵ
:	߮௝ ൌ ߴ௝ ൌ 0 : No Oil prices spillover effects on stock returns
	H଴
ଶ
:	ߛ௝ ൌ ߠ௝ ൌ 0 : No Global US spillover effects on stock returns
H଴
ଷ
:	ߜ௝ ൌ ߰௝ ൌ 0 : No Regional EU spillover effects on stock returns
	H଴
ସ
:	߮௝ ൌ ߜ௝ ൌ ߛ௝ ൌ 0: No mean spillover effects on stock returns
H଴
ହ
:	ߴ௝ ൌ ߠ௝ ൌ ߰௝ ൌ 0 : No volatility spillover effects on stock returns
Table 3: Wald tests for constant spillover model for entire sample period7
from
September 2000 – March 2012)
Wald1 Wald2 Wald3 Wald4 Wald2
CRO 14.797* 46.178* 67.533* 2.201^ 96.515*
(0.001) (0.000) (0.000) (0.086) (0.000)
CZE 31.864* 244.226* 194.387* 5.164* 302.17*
(0.000) (0.000) (0.000) (0.002) (0.000)
HUN 20.727* 242.231* 168.936* 4.199 466.16*
(0.000) (0.000) (0.000) (0.056) (0.000)
POL 39.776* 319.601* 143.008* 7.549* 418.02*
(0.001) (0.000) (0.000) (0.001) (0.000)
ROM 25.480* 56.351* 79.565* 4.217* 133.75*
(0.000) (0.000) (0.000) (0.006) (0.000)
RUS 52.362* 119.239* 28.973* 2.847# 142.51*
(0.000) (0.000) (0.000) (0.037) (0.000)
TUR 9.4069* 69.932* 35.113* 1.163 75.585*
(0.002) (0.000) (0.000) (0.322) (0.000)
UKR 4.9343* 14.665* 21.380* 17.796* 16.151*
(0.008) (0.000) (0.000) (0.000) (0.000)
Note: P values and std. errors are in the parentheses and brackets,whilst *, #, and ^ represent
significance level at 1%, 5% and 10%, respectively.
The presence of significant mean spillovers from the US stock markets is found with
negative coefficients for Poland, Czech Republic and the Ukraine. The weak mean
spillover effect from regional EU market is statistically significant for Romania. Thus,
results show some mean spillover effects from global US, however, the mean spillovers
from regional market can be considered negligible. According to coefficient estimates
7
Wald1, Wald2 ,Wald3 , Wald4 and Wald5 tests are χ2(2) distributed under the aforementioned
null hypothesis.
REVIEW OF ECONOMIC PERSPECTIVES
17
summarized in Table 2, the mean spillover from oil returns significantly influences
stock returns of the Czech Republic, Poland, Romania and Russia, and the Wald tests
are rejected, accordingly. Judged by the scale of the coefficient, the lagged oil return is
found to be highly positive and significant for Russia (߮௢௜௟, 0.180). This finding can be
interpreted by the fact that Russia is the only oil exporting country amongst the eight
European countries investigated. In other words, oil returns seem to highly and
positively drive Russian stock returns. The volatility spillover effects driven by oil
shocks area appeared to be strongly significant and associated with positive parameter
signs for all countries, except for the Ukraine. GJR-GARCH (1, 1) model also captures
asymmetric effects for stock returns8
. The presence of asymmetric effects is found to be
significant only for stock returns of Hungary and Turkey (Table 2). In other words,
some evidences of leverage effect (Ding et. al., 1993), implying that negative shocks
(bad news) tend to influence stock returns more than positive shocks (good news) are
reported only in the cases of Hungary and Turkey.
The effects of volatility spillovers for a country j are evaluated according to variance
ratios namely, VR-US from global factor, VR-EU from regional factor, VR-OIL from
world factor and local VR-OWN are quantified and reported in Table 4. The variance
ratios assess the proportions of unexpected returns in total conditional variance. The US
volatility spillovers are found to be the most dominating factor for conditional variance
of unexpected returns for all investigated countries except for Croatia and Romania.
More specifically, on average 9.4%-26.7% of volatility spillovers are contributed by US
market shocks.
Table 4: Summary statistics for variance ratios for the entire period (from Sep
2000–Mar2012)
CRO CZE HUN POL ROM RUS TUR UKR
VR EU M 22.713 26.716 24.583 16.851 20.055 10.866 13.855 6.367
SD 0.260 0.276 0.261 0.198 0.225 0.160 0.199 0.008
VR US M 16.187 26.798 25.028 24.837 18.707 21.726 24.791 9.492
SD 0.206 0.260 7.077 0.258 0.211 0.244 0.276 0.180
VR OIL M 8.219 7.802 0.070 20.386 9.576 16.932 6.042 3.132
SD 0.145 0.311 0.115 0.219 0.146 0.210 0.107 0.078
VR OWN M 52.880 38.681 43.331 37.923 51.656 50.474 55.310 81.008
SD 0.345 0.311 0.322 0.292 0.325 0.334 0.242 0.277
Note: M for mean and SD for standard deviation.
In contrast, the EU market shocks account on average for 6.3%-25.7% of volatility
spillovers, which is relatively less than global US volatility spillovers. The result aids to
answer the second research question that the US stock returns, thereby the US S&P 500
index has most magnitude effect on national stock markets of European national
markets. Judged by oil volatility spillover coefficients, the highest oil volatility
8
Symmetrical AR(-1) GARCH (1,1) model is also estimated and reported in Table 1 in
Appendix. Judged by the signs and scale of significant coefficients, it should be noted that
symmetrical and asymmetrical models present almost similar results for all eight individual stock
markets.
Volume 13, Issue 1, 2013
18
spillovers effects are found in the case of Russia (VR OIL=16.9%). Finally, purely local
shocks are more pronounced compared to external shocks, and on average for CEE
countries, means of local shocks are relatively small compared to other European
countries. The interpretation of this is more likely due to the fact that CEE countries are
most integrated into the regional, and are considered to be the most highly emerged
stock markets amongst other examined countries
Additionally, I also empirically tested the purely regional EU and global US mean and
spillover effects by excluding the oil price effects9
, and eventually, quantitative
evaluations the variance ratios are computed and reported in Table 5.
Table 5: Summary statistics (Mean and std. dev.) for Variance ratios while
excluding Oil effect over entire period (from September 2000-March 2012)
CRO CZE HUN POL ROM RUS TUR UKR
VR EU Mean (%) 25.569 30.259 27.312 24.342 21.681 16.543 16.557 8.139
Std. dev 0.281 0.299 0.279 0.257 0.256 0.221 0.227 0.158
VR US Mean (%) 18.624 29.593 27.145 31.348 19.894 26.212 30.730 10.936
Std. dev 0.233 0.283 0.200 0.300 0.239 0.281 0.299 0.207
VR OWN Mean (%) 55.805 40.146 45.542 44.276 58.423 57.243 56.578 80.923
Std. dev 0.341 0.324 0.331 0.329 0.333 0.331 0.344 0.276
The assessment by excluding oil effect also confirms the dominating role of US
volatility spillovers over those from EU spillover effects. In other words, the unexpected
returns of shocks on investigated European stock markets are mostly driven by global
US volatility effects and local shocks.
6.2 Sensitivity analysis
In the previous section I have shown that the mean spillover effects from regional EU
stock market are statistically insignificant. This can be interpreted by the existence of
high correlation (0.828) between the US and EU stock market. Taking into account the
reason of such a high correlation, the so called sensitivity analysis is implemented. The
analysis is mainly based onomitting the U.S one period lagged return in estimation
framework by applying redundant variable test. Table 6 reports the results. The findings
indicate that with exclusion of US one period lagged return, the mean spillover effects
from regional market becomes significant for Croatia, Czech Republic, Hungary,
Poland, and the Ukraine. None of the changes are reported for oil mean spillover
effects. Thus, the sensitivity analysis throughout redundant variable tests draws
attention to several brief conclusion that by omitting the US one lagged return yield the
EU mean spillover effects to have strong evidences towards stock returns of individual
European emerging and developing economies.
9
The empirical results and Wald tests are reported in Table 2 and 3 Appendix.
REVIEW OF ECONOMIC PERSPECTIVES
19
Table 6: Estimated results by Exclusion US effect (2000-2012)
δ ߮
Croatia 0.108# 0.025
(0.018) (0.274)
[0.045] [0.023]
Czech 0.097# 0.044^
(0.032) (0.054)
[0.023 [0.055]
Hungary 0.176* 0.045
(0.002) (0.110)
[0.058] [0.028]
Poland 0.144* 0.064#
(0.003) (0.019)
[0.049] [0.027]
Romania 0.108^ 0.080#
(0.053) (0.011)
[0.058] [0.031]
Russia 0.040 0.106*
(0.391) (0.006)
[0.047] [0.039]
Turkey 0.101 0.050
(0.185) (0.310)
[0.076] [0.049]
Ukraine 0.302* -0.070#
(0.000) (0.029)
[0.056] [0.036]
Note: P values and std. errors are in the parentheses and brackets, whilst *,#, and ^ represent
significance level at 1%, 5% and 10%, respectively.
Judging by obtained signs and scales of coefficients in the Table 6, for some countries,
the strong indication of EU mean spillovers using redundant variable test seems to be
interpreted by previous high correlation (0.828) between the US and EU stock returns. It
is worth to note that high correlation between two explanatory variables, namely the US
and EU lagged returns, possibly introduce Multicollinearity problem which cannot be
ignored (Brook, 2008).
6.3 Asymmetric spillover effects on stock returns
Although the asymmetric term is associated in primary spillover models and is found to
be significant for some countries (Hungary and Turkey), the approach mentioned above
sheds more light on exploring the response of stock returns separately towards negative
and positive market shocks and upturns and downturns in lagged returns transmitted to
global (GF US), regional (RF EU), and world factors (WF Oil). Since the mean
specifications for global and regional factors are negligible, the asymmetric effects from
all transmitted market shocks and only from one lagged oil price returns are reported
and analyzed. The Table7 summarized the estimation results for asymmetric tests on
national stock returns. Except for Croatia, all other stock markets respond
Volume 13, Issue 1, 2013
20
asymmetrically to upturns and downturns in oil price, where the estimated coefficient of
ܴ௢௜௟,௧ିଵ
ା
and ߮ଵ is highly significant10
.
ܴ௝,࢚ ൌ ܿ଴,௝ ൅ ܿଵ,௝ܴ௝,௧ିଵ൅	ߛ଴,௝ܴ௎ௌ,௧ିଵ
ି
൅	ߛଵ,௝ܴ௎ௌ,௧ିଵ
ା
൅	ߜ଴,௝ܴா௎,௧ିଵ
ି
൅ ߜଵ,௝ܴா௎,௧ିଵ
ା
൅
߮ଵ,௝ܴ௢௜௟,௧ିଵ
ି
൅ ߮ଵ,௝ܴ௢௜௟,௧ିଵ
ା
൅ ߠ଴,௝߳௎ௌ,௧
ି
൅ ߠଵ,௝߳௎ௌ,௧
ା
൅ ߰଴,௝߳ா௎,௧
ି
൅ ߰ଵ,௝߳ா௎,௧
ା
൅	ߴ଴,௝߳௢௜௟,௧
ି
൅
ߴଵ,௝߳௢௜௟,௧
ା
൅ ߳௝,௧ (19)
Table 7: Estimated coefficients for the asymmetric spillover model, Sep 2000 –
Mar 2012
Croatia Czech Hungary Poland Romania Russia Turkey Ukraine
߮଴ 0.009 -0.054# 0.047 0.072* 0.077# 0.110* 0.049 -0.096
(0.833) (0.020) (0.102) (0.009) (0.015) (0.005) (0.320) (0.007)
[0.045] [0.046] [0.052] [0.050] [0.058] [0.067] [0.082] [0.078]
߮ଵ 0.030 1.092* 1.239* 1.203* 0.923* 0.701* 1.009* 0.301*
(0.528) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
[0.030] [0.047] [0.067] [0.060] [0.065] [0.068] [0.100] [0.080]
߰଴ 0.872* 0.780* 0.866* 0.979* 0.605* 0.782* 1.046* 0.280*
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
[0.132] [0.089] [0.108] [0.115] [0.126] [0.165] [0.214] [0.137]
߰ଵ 0.636* 0.179* 0.202* 0.231* 0.227* 0.364* 0.212* 0.069
(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.175)
[0.133] [0.122] [0.142] [0.139] [0.180] [0.210] [0.224] [0.215]
ߠ଴ 0.554* 0.379* 2.589* 0.341# 0.389^ 0.499* 0.189# 0.947*
(0.000) (0.005) (0.001) (0.039) (0.052) (0.006) (0.050) (0.002)
[0.072] [0.056] [0.061] [0.066] [0.100] [0.090] [0.156] [0.113]
ߠଵ 0.248* 0.125* 0.063 0.115* 0.104* 0.122* 0.029# 0.236*
(0.008) (0.000) (0.215) (0.003) (0.000) (0.002) (0.038) (0.000)
[0.093] [0.081] [0.094] [0.091] [0.141] [0.123] [0.171] [0.157]
ߴ଴ 0.086 -0.027 0.161# -0.005 -0.020 0.001 0.058* 0.067
(0.087) (0.468) (0.025) (0.893) (0.454) (0.998) (0.001) (0.287)
[0.057] [0.042] [0.057] [0.050] [0.072] [0.067] [0.096] [0.086]
ߴଵ 0.181* 0.836* 0.619* 0.853* 0.887* 0.849* 0.936* 0.767*
(0.002) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)
[0.048] [0.061] [0.060] [0.056] [0.064] [0.087] [0.104] [0.096]
Note: P values and std. errors are in the parentheses and brackets, whilst *,#, and ^ represent
significance level at 1%, 5% and 10%, respectively.
The coefficients of R୭୧୪,୲ିଵ
ି
and φ଴ are found to be significant for the Czech Republic,
Poland, Romania and Russia. In other words, it seems that the increase in oil returns has
a greater effect rather than the decrease, and the magnitude of increase is greater for
countries such as Poland, Czech Republic and Russia. The Wald test (H଴
ହ
:	φ଴ ൌ φଵ) of
no oil shocks, however, is rejected for Romania and the Ukraine only. The results in
Table 7 also indicate that stock markets of all the European emerging and developing
countries except for Croatia respond asymmetrically to oil shocks. In spite of that,
Romania, Russia and the Ukraine seem to respond significantly only to positive shocks,
however, the stock returns of the Czech Republic, Hungary, Poland, and Turkey
significantly depend on decrease and increase oil commodity oil price. The evidence is
10
Note: the constant one period own lagged of individual countries are not reported in estimation
results.
REVIEW OF ECONOMIC PERSPECTIVES
21
that stock returns of individual countries tend to asymmetrically react to oil price
shocks.
Furthermore, some indication of asymmetric existences towards global US and regional
EU market shocks is recorded, too. Strong evidences of asymmetric response for US
and EU contemporary residuals specified with negative (߳௎ௌ,௧
ି
and߳ா௎,௧
ି
ሻ and positive
൫߳௎ௌ,௧
ା
	and	߳ா௎,௧
ା
൯ shocks are obtained. Based on the significant coefficient values, the
negative shocks in both US and EU markets have more effects than positive shocks,
implying that investors tend to react sharply to the negative shocks rather than the
positive ones.
Despite the fact that regression coefficients appeared strongly for contemporary
residuals from oil market as well as the US and EU stock markets, the joined Wald tests
are not rejected in most of the cases. More specifically, in the case of regional EU
shocks, the joined Wald tests of no asymmetric responses to EU shocks are rejected
only for Hungary, Poland, and Russia at 5% significance level. Not surprisingly,
asymmetric responses towards negative shocks are more substantial thanks to positive
shocks caused by regional EU residuals in the cases of Hungary, Poland, and Russia. In
addition, Russia is the only market to respond significantly to oil price shocks, yet at
10% significance level. Moreover, the asymmetric shocks from US market are reported
to be jointly significant only in the cases of Croatia and Romania. Taking into account
the empirical findings fromWald test results, some evidences of asymmetric effects can
be concluded (Table 4 in Appendix). However, the stock returns of the European
emerging and developing markets do not react asymmetrically to either contemporary
residuals or returns in most of the cases.
6.4 Conditional spillover model results
In order to evaluate performance of spillover intensities caused by global and regional
stock market, conditional spillover modelspecified by equations (20)-(23) are relevantly
explored for eight individual European countries. For investigation of behavior of EU
and US spillovers, instrumental variables such as exchange rates and total trade/GDP
are applied.
The results shown in Table 8 indicate that total trade to GDP ratio with positive
coefficient (0.973) is the only significant variable explaining the EU conditional
spillover effects for Croatia. Similar arbitrary empirical results are obtained for the
Czech Republic and Romania. More specifically, in the case of the Czech Republic, US
conditional mean spillovers and EU volatility spillovers are significantly driven by
exchange rate changes, with coefficients of -0.630 and 1.122.
On the other hand, for Romania, the US conditional mean spillover effect is explained
by total trade/GDP ratio with corresponding positive coefficient of 0.159. Moreover,
both US conditional mean and volatility spillovers are not found to be strong in the case
of Hungary and Poland. The exchange rate change is also a strong proxy in explaining
the behavior of EU conditional volatility transmissions for Poland, while the total
trade/GDP is a strong proxy in the case of Hungary. For both Hungary and Poland,
Volume 13, Issue 1, 2013
22
conditional mean spillovers from regional markets are reported to be significantly
introduced by two macroeconomic instrumental variables. Still more interesting is the
case of Russia, both EU conditional mean and volatility spillovers are not found to be
driven by constructed instruments. However, the estimated instruments for explaining
the parameters of US conditional mean and volatility spillover effects are found to be
highly significant and positive. There is strong empirical evidence that US conditional
spillover intensities can be explained by exchange rate fluctuations of USD and ruble, as
well as by the total trade between US and Russia. When it comes to findings for Turkey,
only exchange rate changes, namely LIRA/USD and LIRA/EURO can significantly
impact the US and EU conditional spillover parameters.
Table 8: Estimated coefficients for the conditional spillover model, from Sep 2000
to Mar 2012
CRO CZE HUN POL ROM RUS TUR UKR
US conditional Mean spillover
Constant v0 2.555 -1.668 5.520 1.140 4.446 -4.982 0.835 -6.972^
(0.352) (0.724) (0.196) (0.819) (0.156) (0.140) (0.750) (0.077)
[2.748] [4.731] [4.271] [1.140] [3.136] [3.376] [2.627] [3.957]
Exchange rate v1 -0.774 -0.630# -0.497 -0.097 1.199 2.916* 1.782* 1.683#
changes (0.108) (0.036) (0.324) (0.872) (0.080) (0.004) (0.000) (0.023)
[0.482] [0.300] [0.504] [0.608] [0.687] [1.005] [0.363] [0.742]
Trade to GDP ratio v2 0.084 0.353 0.241 0.071 0.159# 0.450* 0.119 -0.340
(0.728) (0.405) (0.483) (0.772) (0.044) (0.003) (0.635) (0.136)
[0.244] [0.424] [0.344] [0.460] [0.298] [0.123] [0.252] [0.228]
EU conditional Mean spillover
Constant p0 -0.061 4.037 -33.459# 3.606 0.039 1.847 0.305 -3.212
(0.962) (0.648) (0.019) (0.523) (0.993) (0.794) (0.936) (0.452)
[12.89] [8.861] [14.35] [5.652] [4.692] [7.076] [3.807] [4.273]
Exchange rate p1 4.239 -0.921 4.656# 3.563* 1.447 0.210 1.026* 0.839
changes (0.529) (0.366) (0.024) (0.003) (0.140) (0.884) (0.004) (0.186)
[6.739] [1.021] [2.074] [1.234] [0.982] [1.443] [0.361] [0.634]
Trade to GDP ratio p2 0.973# 0.084 0.982 0.132 0.215 0.291 0.063 -0.207
(0.021) (0.937) (0.136) (0.840) (0.655) (0.277) (0.880) (0.586)
[0.432] [1.083] [0.659] [0.655] [0.482] [0.268] [0.420] [0.362]
US conditional Volatility spillover
Constant w0 2.094 -0.951 3.781 1.180 3.783 -6.278^ 1.239 -8.485#
(0.456) (0.846) (0.384) (0.813) (0.233) (0.064) (0.648) (0.029)
[8.810] [4.910] [4.351] [5.008] [3.173] [3.395] [2.721] [3.897]
Exchange rate w1 -0.659 -0.540^ -0.338 0.115 1.264^ 3.206* 1.735* 1.933*
changes (0.180) (0.079) (0.514) (0.854) (0.067) (0.001) (0.000) (0.008)
[0.492] [0.308] [0.519] [0.630] [0.690] [1.001] [0.377] [0.736]
Trade to GDP ratio w2 0.067 -0.256 0.165 0.101 0.539^ 0.424* 0.161 -0.426^
(0.787) (0.565) (0.636) (0.827) (0.076) (0.000) (0.537) (0.060)
[0.250] [0.446] [0.350] [0.462] [0.304] [0.124] [0.261] [0.225]
REVIEW OF ECONOMIC PERSPECTIVES
23
EU conditional Volatility spillover
Constant q0 -6.200 -6.962 -13.691 -3.805# -2.621 4.682* -0.817 -1.010
(0.389) (0.131) (0.123) (0.021) (0.311) (0.002) (0.733) (0.674)
[7.205] [4.612] [8.889] [3.099] [2.590] [4.357] [2.405] [2.409]
Exchange Rate q1 2.696 1.122# 1.319 1.316# 0.178 -1.060 -0.407# -0.172
Changes (0.479) (0.046) (0.341) (0.043) (0.747) (0.239) (0.045) (0.656)
[3.809] [0.563] [1.385] [0.651] [0.555] [0.902] [0.202] [0.388]
Trade to GDP ratio q2 -0.104 0.401 -0.800# -0.231 -0.319 0.112 -0.114 -0.153
(0.627) (0.483) (0.020) (0.510) (0.228) (0.476) (0.633) (0.442)
[0.215] [0.572] [0.346] [0.351] [0.265] [0.158] [0.263] [0.199]
Note: P values and std. errors are in the parentheses and brackets, whilst *,#, and ^ represent
significance level at 1%, 5% and 10%, respectively.
Finally, none of the time varying spillover models can be explained by exchange rate
changes and total trade to GDP ratio in the case of the Ukraine. The insignificant
evidences for the Ukraine draws attention to exploration of other macroeconomic or
financial data to further empirically investigate conditional spillover effects.
The overall overview is that instrumental variables, namely exchange rate changes and
Trade/GDP ratios, do not properly introduce clear cut results for all the countries to
explain conditional spillover models. It should be noted, however, that among the
obtained empirical results for instrumental intensities, exchange rate changes seem to
me more powerful instrument in explaining the US and EU spillover parameters rather
than trade/GDP ratio.
Finally, in order to fully evaluate the conditional model, variance ratios are also
computed. The Table 9 shows that similar to constant model volatility spillover effects
local shocks account for most of the market shocks on total variance, and the US
variance ratios are still substantial compared to the EU and oil ratios.
Table 9: Summary statistics forVariance ratios for conditional spillover mode over
the entire period, (from Sep 2000 to March 2012)
CRO CZE HUN POL ROM RUS TUR UKR
VR EU Mean 21.117 24.469 23.240 20.168 16.716 9.456 12.122 8.553
Std.dev 0.249 0.266 0.256 0.232 0.214 0.143 0.182 0.160
VR US Mean 16.118 26.541 24.913 27.777 17.745 21.047 25.102 10.136
Std.dev 0.205 0.252 0.257 0.275 0.207 0.239 0.279 0.182
VR OIL Mean 8.458 8.183 7.137 7.967 9.154 17.169 6.142 2.076
Std.dev 0.149 0.140 0.116 0.124 0.157 0.218 0.118 0.065
VR OWN Mean 54.246 40.665 44.708 44.093 56.383 52.035 56.633 79.230
Std.dev 0.345 0.319 0.322 0.322 0.335 0.335 0.242 0.286
Volume 13, Issue 1, 2013
24
Similar to constant spillover model, the volatility spillovers effects from the US market
shocks account for the most of the variances compared to regional EU stock market and
oil market in conditional model. In contrast to constant spillover model, the regional and
global volatility spillovers are almost equally distributed for the Czech Republic,
Hungary, Romania, Poland, and the Ukraine. Eventually, own shocks appeared greatly
in the case of the Ukraine which was then followed by Turkey and Croatia.
6.5 AR (1)-GJR-GARCH (1, 1) dummy model: EU enlargement effect
By using the equation (24) given by GARCH model, the effect of EU accession is
empirically examined for EU member countries, namely for the Czech Republic,
Hungary, Poland and Romania. Table 10 shows that significant results are obtained only
in the case of Romania upon admission of the country into EU.
Table 10: Dummy variable model estimation results11
Czech Hungary Poland Romania
c0 0.583* 0.591# 0.147 0.629*
(0.000) (0.013) (0.497) (0.001)
[0.174] [0.238] [0.217] [0.202]
c1 -0.073^ -0.064 -0.106# -0.031
(0.085) (0.139) (0.021) (0.466)
[0.042] [0.043] [0.043] [0.043]
ߜ -0.032 0.121 -0.029 0.184#
(0.642) (0.127) (0.714) (0.047)
[0.070] [0.079] [0.081] [0.093]
γ 0.173# 0.083 -0.235* -0.110
(0.012) (0.293) (0.005) (0.276)
[0.069] [0.078] [0.085] [0.101]
ߴ 0.057# 0.050^ 0.072 0.082#
(0.017) (0.079) (0.009) (0.010)
[0.023] [0.028] [0.028] [0.031]
Ф 1.083* 1.240* 1.205* 0.903*
(0.000) (0.000) (0.000) (0.000)
[0.057] [0.067] [0.071] [0.077]
ФDum -0.328 -0.455 0.078 -0.555#
(0.119) (0.112) (0.753) (0.050)
[0.210] [0.286] [0.250] [0.286]
ߠ 0.774* 0.871* 0.979* 0.608*
(0.000) (0.000) (0.000) (0.000)
[0.035] [0.039] [0.041] [0.056]
߮ 0.182* 0.206* 0.231* 0.228*
(0.000) (0.000) (0.000) (0.000)
[0.023] [0.031] [0.027] [0.034]
11
The Table 16 presents results only for mean equation in GJR-GARCH model.
REVIEW OF ECONOMIC PERSPECTIVES
25
Note: PP values and std. errors are in the parentheses and brackets, whilst *,#, and ^ represent
significance level at 1%, 5% and 10%, respectively.
More specifically, the findings indicate that the admission of Romania in EU in 2007
seems to be statistically significant, still with a negative coefficient of -0.555. In other
words, in the case of Romania volatility spillovers seem to decrease after the EU
accession. However, the results of EU enlargement effect are significant in 2004.
Additionally, quantified variance ratios for GARCH dummy model also revealed that
the proportion of variance caused by regional EU effect is relevantly small for Romania,
which is 9.3%. Furhemore, quantified variance ratios for dummy Garch model is
presented below in Table 11.
Table 11: Summary statistics for variance ratios for dummy variable model period
from September 2000 to March 2012
CZE HUN POL ROM
VR EU Mean 18.566 15.005 23.105 9.335
Std. dev 0.229 0.199 0.250 0.118
VR US Mean 29.469 28.400 27.536 20.385
Std. dev 0.272 0.235 0.275 0.246
VR OIL Mean 9.221 8.593 7.593 11.492
Std. dev 0.138 0.136 0.122 0.164
VR OWN Mean 42.742 48.000 41.764 58.787
Std. dev 0.316 0.332 0.318 0.317
7. Conclusion
The paper studied the mean and volatility spillover effects from US, and EU stock
markets, as well as from the oil market to eight individual European stock markets.
Applying GJR-GARCH model, I found strong evidences of volatility transmission,
particularly global, regional and world factors towards the national stock markets of
eight European countries. The empirical outcomes also showed that amongst the three
external factors, the US volatility spillover intensities account for most of the proportion
of unexpected returns, except for Croatia and Romania, though. The empirical findings
are also similar for pure global and regional stock markets while excluding the world
factor oil. The empirical results of mean spillover effects are mixed and imply no strong
evidences for Croatia, Hungary and Turkey. In addition to that, through various
specifications in the so-called sensitivity analysis, I have revealed that the EU mean
spillover effects are fairly sensitive in conjunction with the US mean spillover effects
towards individual stock markets countries. Moreover, the results also showed that for
no European Union member country is highly influenced by their own local shocks,
which , however, appeared to be the highest in the Ukraine that was followed by Turkey
and Croatia.
Furthermore, oil market shocks are found to be significant for all countries and, in
particular, drive the stock returns of Russia with very high and positive coefficients.
Volume 13, Issue 1, 2013
26
This finding is readily explained by a higher presence of oil and gas sector companies in
the total market capitalization for Russian stock market. On the other hand, there are
weak indications of asymmetric responses. More specifically, only Romania, Poland
and the Ukraine asymmetrically responded to EU market shocks. Asymmetrical
responses towards US shocks are found only in the case of Romania. Only the stock
returns of Russia respond asymmetrically to oil price market shocks, though the
response is fairly weak.
Additionally, in order to test the European Union enlargement effect through spillover
effects on stock markets, I utilized the presence of a dummy variable constant model. I
found out that the effect is evident only in the case of Romania. Thus, the overall
inference of the dummy model is that EU membership matters for stock returns in
Romania while being significant and still negative. Although the significance level is
not statistically high, it can reflect some evidence of hypothesis on EU enlargement
effect.
Finally, I also found statistically significant results for a conditional model and the
conditional model appeared to be prior to the constant spillover model. Overall, the
empirical outcomes of conditional spillover model can be summarized based on two
essential inferences, the first of which is related to estimation results on parameters of
global and regional markets. More specifically, empirical findings of exchange rate
changes in most cases are highly significant and positive for US spillover effects, while
when judged by sign and scale of coefficients for the EU spillover effects, the empirical
results are found to be relatively weak or insignificant. The second implication is that
most of the parameters both for mean and spillover effects are significantly explained
by exchange rate changes rather than the total trade/GDP ratio, which shows the relative
importance of exchange rate fluctuations for spillover effects amongst examined
European countries.
Reference
BAELE, L. (2005). Volatility Spillover Effects in European Equity Markets, Journal of
Financial and Quantitative Analysis 40(2), Pp. 373–401.
BEINE, J., CAPORALE, G.M., GHATTAS, M. S., SPAGNOLO, N. (2010). Global
and regional spillovers in emerging stock markets: A multivariate GARCH-in-mean
analysis. Emerging Markets Review, 11(2) Pp. 250–260.
BEKAERT, G., HARVEY, C.R. (1997). Emerging Equity Market Volatility, Journal of
Financial Economics, 43(1), Pp. 29–77.
BOLLERSLEV, T. (1986). Generalized Autoregressive Conditional Heteroskedasticity,
Journal of Econometrics 31(3), Pp. 307–327.
BROOKS, C., (2008). Introductory Econometrics for Finance, Cambridge, University
press, Pp.170-174.
CHRISTIANSEN, C. (2004). Decomposing European Bond and Equity Volatility,
Working paper, F -2004-01 Finance Research Group, Aarhus School of Business.
REVIEW OF ECONOMIC PERSPECTIVES
27
GILMORE, C. G., LUCEY, B. M., MCMANUS, G. B., (2006). The dynamics of
Central European equity market comovements” The Quarterly Review of Economics
and Finance, vol.48 (1) Pp. 605–622. DOI:10.1016/j.qref.2006.06.005.
ENGLE, R. F. (1982). Autoregressive Conditional Heteroskedasticity with Estimates of
the Variance of United Kingdom Inflation, Econometrica, vol. 50 (4), Pp., 987–1007.
ENGLE, R. F., NG, K.V. (1993). Measuring and testing the impact of news on
volatility. The Journal of Finance, Vol.48, (5), pp. 1749–1778. DOI: 10.1111/j.1540-
6261.1993.tb05127.x.
GOETZMANN, W., IBBOTSON, G.R. (2006). The equity risk premium: Essays and
explorations. Oxford University Press, pp. 368–367.
HAMAO, Y., MASULIS, R., NG, K.V. (1990). Correlation in prices changes and
volatility across international stock markets, The Review of Financial Studies 3(2) Pp.
281–307.
KASMAN, A., KASMAN, S., TORUN, E. (2009). Dual long memory property in
returns and volatility: Evidence from the CEE countries’ stock markets, Emerging
Markets Review. 9(4) Pp.122–139. DOI: 10.1016/j.ememar.2009.02.002.
KOCENDA, E., HANOUSEK, J. (2010). Foreign News and Spillovers in Emerging
European Stock Markets, William Davidson Institute Working Paper, 983.
GILMORE C.G., MCMANUS, M.G. (2002). International portfolio diversification: US
and Central European equity markets, Emerging Markets Review.3(1), pp. 69–83. DOI:
10.1016/S1566-0141(01)00031-0.
ÉGERT, B., KOUBAA, Y. (2004). Modelling Stock Returns in the G-7 and in Selected
CEE economies: A Non-linear GARCH Approach. William Davidson Institute Working
Paper, No. 663.
DVORAK, T., PODPIERA, R. (2006). European Union Enlargement and Equity
Markets in Accession Countries. Emerging Markets Review, Vol. 7 (2), Pp.129–146.
NG, A., (2000). Volatility Spillover Effects from Japan and the US to the Pacific-Basin.
Journal of International Money and Finance, 19 (2) Pp. 207–233. DOI: 10.1016/S0261-
5606(00)00006-1.
ROCKINGER, M., URGA, G. (2001). The evolution of stock markets in transition
economies. The Journal of Business & Economic Statistics, vol. 19, pp.73–84. DOI:
10.1006/jcec.2000.1669.
WALTER, E. (2010). Applied Econometric times Series. John Wiley & Sons, Inc.,
pp.155–156.
Volume 13, Issue 1, 2013
28
Appendix
Table A2: Wald test results in asymmetric model from September 2000 to March
2012
Wald1 Wald2 Wald3 Wald4 Wald5 Wald6
CRO 2.936^ 1.618 0.681 1.089 4.857 1.309
(0.087) (0.203) (0.794) (0.297) (0.057) (0.252)
CZE 0.023 1.175 0.108 1.402 0.065 0.377
(0.878) (0.278) (0.741) (0.236) (0.798) (0.539)
HUN 0.386 0.383 0.493 5.439# 0.038 0.001
(0.584) (0.536) (0.482) (0.020) (0.845) (0.986)
POL 0.665 0.012 0.266 4.679# 0.554 1.429
(0.415) (0.910) (0.606) (0.030) (0.456) (0.232)
ROM 1.318 0.002 3.120^ 2.204 7.358* 0.446
(0.251) (0.962) (0.077) (0.138) (0.006) (0.504)
RUS 0.318 0.057 0.001 3.741# 0.001 5.561^
(0.572) (0.810) (0.991) (0.050) (0.981) (0.090)
TUR 0.104 0.808 0.409 0.432 0.001 1.607
(0.746) (0.369) (0.522) (0.511) (0.996) (0.205)
UKR 1.701 0.402 2.515 17.300* 1.317 0.650
(0.192) (0.526) (0.078) (0.000) (0.251) (0.420)
Note: P values and std. errors are in the parentheses and brackets, whilst *,#, and ^ represent
significance level at 1%, 5% and 10%, respectively, and also null hypotheses of joined Wald tests
are presented below.
Table A3: Wald test results under oil exclusion constant spillover model
(September 2000- March2012)
Wald1 Wald2 Wald3 Wald4
CRO 52.562* 69.398 3.547 121.11
(0.000) (0.000) (0.029) (0.000)
CZE 206.07 190.57 7.259 379.82
(0.000) (0.000) (0.002) (0.000)
HUN 201.93 163.76 6.395 570.30
(0.000) (0.000) (0.002) (0.000)
POL 238.27 131.83 9.873 376.69
(0.000) (0.000) (0.001) (0.000)
ROM 61.385 87.802 4.851 171.67
(0.000) (0.000) (0.009) (0.000)
RUS 103.73 29.021 4.158 138.54
(0.000) (0.000) (0.021) (0.000)
TUR 74.022 39.093 2.236 108.87
(0.000) (0.000) (0.108) (0.000)
UKR 15.426 24.110 10.048 42.931
(0.000) (0.000) (0.001) (0.000)
REVIEW OF ECONOMIC PERSPECTIVES
29
Note: P values and std. errors are in the parentheses and brackets, whilst *,#, and ^ represent
significance level at 1%, 5% and 10%, respectively, and also null hypotheses of joined Wald tests
are presented below:
‫ܪ‬଴
ଵ
:	ߛ௝ ൌ ߠ௝ : no global US spillover effects
‫ܪ‬଴
ଶ
:	ߜ௝ ൌ ߰௝ : no regional EU spillover effects
‫ܪ‬଴
ଷ
:	ߛ௝ ൌ ߜ௝ : no mean spillover effects
‫ܪ‬଴
ସ
:	ߠ௝ ൌ ߰௝ : no asymmetric responses to US shocks