Industrial and Financial Economics Master Thesis No 2004:66 Nordic Financial Market Integration: An Analysis with GARCH Modeling

Industrial and Financial Economics Master Thesis No 2004:66

Nordic Financial Market Integration:

An Analysis with GARCH Modeling

Ahou Virginie Baudouhat

Graduate Business School

School of Economics and Commercial Law Göteborg University

ISSN 1403-851X

Printed by Elanders Novum

To my parents Maria Elena and Ngoran Dominique for inspiring me to have dreams and to have the courage to live them.

To my brothers Jean François and Max Jacob for their care and support of every instant.

Acknowledgements

I would like to express my greatest gratitude to Lennart Flood for his precious supervision and guidance and to Jianhua Zhang for his valuable and skillful comments on my work.

I am also grateful to the Center for Finance of Göteborg Graduate Business School for providing the access to the study database and for insightful suggestions.

Finally, I express my special thanks to Ann Veiderpass, who followed my

work with keen interest and intellectual criticism and who is a source of

inspiration for my future.

Contents

Abstract...6

1 Introduction ...7

2 Empirical Specification ...10

2.1. Model Specification...10

2.2. Explanatory Variables ...15

3 Data...16

4 Empirical Results...21

4.1. Predictability of Market Index Returns with the Explanatory Variables ...21

4.2. Some evidence on Nordic Market Integration ...23

4.3. Financial Integration over Time ...25

5 Conclusion...28

References ...30

Tables...32

Abstract

This thesis investigates the financial integration of the Nordic stock markets by studying the return-spillover effects across countries.

Three related hypotheses are addressed. Firstly, the increasingly documented dominance of the EMU market over the US market is assumed to apply for the Nordic countries. Secondly, the non-EMU members are expected to be more integrated with each other than with the Finnish participating country. Corresponding expectations of a higher and exclusive integration of Finland with the EMU than with any of the Nordic markets are formed. Thirdly, the Euro introduction is assumed to have had a significant impact on the integration of all the Nordic markets.

Using a GARCH(1,1) model, we analyze the degree and evolution through time of integration between the Nordic stock markets. It is found that the first hypothesis cannot be rejected as the EMU is the dominant market for the Nordic countries. The second hypothesis cannot be rejected for Denmark and Norway. Sweden is found to be more integrated with Finland than with Denmark and Norway. Finally, the third hypothesis is rejected as there is evidence that financial market integration in the Nordic region has not been significantly influenced by the European unification process.

Keywords: financial integration, stock markets, EMU, Nordic countries,

GARCH, volatility.

1 Introduction

On January 1, 1999, eleven European Union (EU) member countries implemented a common currency, the Euro. The adoption of the Euro is a milestone in regional economic integration and is seen by many authors as

“the most dramatic development in international finance since the collapse of the Bretton Woods system” (Yang J. et al., 2004). The structural environment of European financial markets has significantly changed over the past two decades, through deregulation, the opening to foreign-owned intermediaries and growing electronic trading. During the same time, the international integration of financial markets around the world has dramatically increased. Despite the springboard implemented throughout the Euro launch, the creation of a single financial market in Europe is still in process and Europe’s diverse and separate equity markets are moving towards closer cooperation with relevant developments such as broadening privatization and an increasing number of pan-European mergers and acquisitions.

As underlined by Adler and Dumas (1983), there is no precise definition of

capital market integration in the current literature. Integration is generally

opposed to segmentation depending on whether or not barriers to

investment exist between countries. Barriers to investment are essential

factors that can prevent markets to integrate. Among these factors are

exchange rate risk, legal and tax differences, information availability,

foreign ownership restrictions, homes bias (Stulz, 1981; Errunza and Losq,

1985). In perfectly integrated markets all assets with identical risk exposure

also command identical expected returns (Campbell and Hamao, 1992). In

addition, a high degree of integration between national markets minimizes

the potential benefit from international diversification (Bessler and Yang,

2003).

Following the seminal work by Errunza, Losq and Padmanabhan (1992) and Errunza and Losq (1985), international equity market integration has been noticeably developed in the literature and numerous methodologies for its measurement have been applied. An important part of the literature focuses on asset pricing models. These include studies of a world CAPM (Harvey, 1991), a world CAPM with exchange risk (Dumas and Solnik, 1995), a world consumption-based model (Wheatley, 1988), world arbitrage pricing theory (Solnik, 1983), world multibeta models (Ferson and Harvey, 1993), a model in between segmentation and integration - the so-called mild segmentation model (Errunza, Losq, and Padmanabhan, 1992), a CAPM with time-varying coefficients (Bekaert and Harvey, 1995).

Correlation and covariance matrix estimates have been used by Longin and Solnik (1995). Other authors have conducted their research on financial market integration through a generalized ARCH (GARCH) structure such as Hamo et al. (1990), Fratzscher (2001).

Even though much of the existing literature is mainly focused on countries and areas other than Europe, there has been an increased interest in financial integration in Europe especially with regards to the launch of the EMU.

The introduction of the euro has contributed to the creation of an integrated financial market in the EU, notably through the suppression of the principal obstacles to the performance of financial services across borders. There is ample evidence that full convergence in European bond markets and money markets have been achieved (Frankel, 1994). However, recent studies state that the level of integration accessed in the different compartments of the European equity market is still very heterogeneous.

In addition, significant barriers and obstacles have the effect to create

market segmentation and result from the differences between EU

jurisdictions: the diversity of legal systems of the member states, the

disparity of enforceable legislation, difference in taxation systems and also cultural diversity.

Regulators in the Nordic countries agree that substantial progress of integration has been achieved during the last decade as well as increasing efficiency of financial markets in the region. However, they emphasize the need for further integration as this could deliver substantial economic benefits and advance development of all sectors of the economy. The Nordic region has a specific position within Europe as regards the differences among its countries in their relationship to the EMU. Indeed, only Finland has joined the EMU so far, and Denmark, Norway and Sweden remain outside the EMU. In addition, there is no categorical scheme towards adopting the euro for the time being. At the dawn of the euro introduction, there was a generally positive attitude towards the single currency and Nordic financial markets believed that the introduction of the euro would create new trading opportunities irrespective of their participation to the single currency.

The purpose of this thesis is to investigate the financial integration of the Nordic stock markets by looking at the return-spillover effects across countries. We address three related hypotheses.

•

H1. There is empirical evidence on the US market being the dominant market in explaining returns on any given national stock market worldwide. However, the recent literature on financial integration in Europe has proved the domination of the EMU market for European financial markets. Accordingly, we expect our results to confirm this feature.

•

H2. In relation to the participation to the single currency, we expect

the non-EMU members to be more integrated with each other than

expectations of a higher and exclusive integration of Finland with the EMU, than with any of the Nordic markets.

•

H3. We stress that the Euro introduction has had a significant impact on the integration of all the Nordic markets.

Our measure of capital market integration is based on the one developed by Fratzscher (2001) who uses a GARCH framework. Applying a GARCH(1,1) model, we analyze the degree and evolution through time of integration between the Nordic stock markets.

The rest of the thesis is organized as follows. Section 2 outlines the empirical specification. Section 3 describes the data and the explanatory variables. Section 4 presents the main empirical results. Section 5 concludes.

2 Empirical Specification

2.1. Model Specification

A common observation about the unexpected component of asset returns

is that large shocks tend to be followed by larger shocks, and small shocks

tend to be followed by more small changes, in either direction. In other

words, the volatility of asset returns appears to be serially correlated. The

econometric term describing this feature is the autoregressive conditional

heteroscedasticity (ARCH), which states that the variance of time series is

conditional on their past realizations. The standard ARCH model was

introduced by Engle (1982) and generalized (GARCH) by Bollerslev

(1986). In his model, Engle defines the conditional variance as a

deterministic function of lagged squared residuals. In the ARCH(q) model

the conditional variance is given by

2 0 1

2

j j t q

t

^{= α +} ∑

^{α ε}

σ

₍₁₎

Bollerslev extend Engle’s specification by introducing lagged conditional variances in the conditional variance equation. This representation allows the number of parameters in the model to be considerably reduced. The GARCH model is commonly used in its most simple form, the GARCH(1,1) model, in which the conditional variance is given by

2 1 1 2

1 1 0

2

= +

+

t

α α ε β σ

σ ₍₂₎

To illustrate the characteristics of ARCH effects, Figures 1 and 2 plot the daily returns on the stock market index for Sweden and Finland. Evidence of an ARCH effect in the markets is seen through the clusters of large positive alternatively negative shocks, for example in the mid 1998’s, and from the mid 1999’s to the mid 2002 (up to the mid 2003 for the Finnish market). Similarly, there are clusters of small positive, alternatively negative, shocks at the two extremes of the sample period.

Figure 1 Return on the Swedish Market Index

Figure 2 Return on the Finnish Market Index

Our specification is based on the idiosyncratic shocks, expressed as the unexpected part (residuals) of the returns of each index series. A two-step estimation is therefore used in order to obtain the residuals first, and then to include their effect in the return on the individual markets.

As a first step, we model the return on each market index by a GARCH(1,1) with no regressors in the variance equation. The mean return on the respective index series is expressed as

t i t

i i i

t

i

R

R

= α

+ α

+ ε

(3)

Where i=Denmark, Finland, Norway, Sweden, EMU, US, R

is the contemporaneous logarithmic return on country i index, R

is the logarithmic return on country i index for the previous period, and ε

is the residual namely idiosyncratic shock for country i index.

The idiosyncratic shocks for Denmark, Finland, Norway and Sweden, as well as the EMU and the US are respectively computed. They represent the sources of return-spillover effects to be subsequently included in the mean equation for the return on each individual Nordic country.

The conditional variance of the index return series also follow a GARCH(1,1) process

2 1 , 2

1 , 2

,_t

=

+

+

i

ω α ε β σ

σ ₍₄₎

The conditional variance of the index series is defined to be dependent on

its own forecasted variance from the last period, σ

and from

information about volatility observed in the previous period, ε

.

Necessary conditions for the conditional variance to be positive is to impose that ω

>0 and α

, β

≥0 (these are referred to as Bollerslev’s conditions). In addition, stationarity of the GARCH process requires that α

+β

≤1, according to Bollerslev’s theorem 1. We preliminary assume that these conditions are fulfilled.

The second step of the model specification allows us to define the return on each market index. The GARCH(1,1) model for the mean return on the individual market index is estimated as

1 , ,

, ,

, , 1

, ,

,^t

=

+

Χ

+ ∑

+

+

R

µ β λ ε λ ε λ ε ₍₅₎

Where i is the individual Nordic market and j represents all other Nordic markets. Χ

is the information set of variables. The idiosyncratic innovations ε

, ε

, and ε

from the previous step are included as explanatory variables in the mean equation for the return on the individual market index. They are assumed to be independent and are intended to capture the degree of market integration in terms of return-spillover effects. US shocks are expressed to affect Nordic markets only on the following day, in order to account for non-synchronous trading between the American and the European markets. The Nordic markets are assumed to operate at approximately the same trading hours.

Due to heteroscedasticity in the residuals ε

, ε

, and ε

the conditional variance is modeled as

2 1 , ,

2 , ,

2 , , 2

1 , 2

1 , 2

,^t

=

+

+

+ ∑

+

+

i

ω β σ α ε α ε α ε α ε

σ ₍₆₎

Equation (6) implies that the conditional variance is determined by its own past variance, and its own past squared shock. In addition, the squared residuals from the first step are included in the variance equation so as to gauge external volatility spillovers to the individual Nordic market. The conditional variance of country i index is therefore presumed to also be affected by outer sources of volatility shocks (from the Nordic neighbor markets, from the EMU and from the US). Innovations from the US are again lagged one period as they are assumed to affect the Nordic markets on the following calendar day.

Equation (5) estimates the dependence of country i index return on shocks generated in the neighbor Nordic stock markets, in the EMU, and in the US. The parameters λ

, λ

and λ

provide our measures for financial market integration.

Equation (6) estimates the proportion of the variance on country i market index that is explained by shocks emanating from the other Nordic markets, as well as from the EMU and the US. The parameters α

, α

and α

are our measures for volatility-spillover effects.

GARCH specifications are estimated by the method of maximum likelihood under the assumption that the errors are conditionally normally distributed. However, violation of the variance fundamentals is likely as we suspect that the residuals are in fact not conditionally normally distributed.

When the assumption of conditional normality does not hold, the ARCH

parameter estimates will still be consistent, provided the mean and

variance functions are correctly specified. However, the estimates of the

covariance matrix will not be consistent resulting in incorrect standard

errors. Therefore, the quasi-maximum likelihood (QML) covariances and

standard errors are computed using the methods of Heteroscedasticity

Consistent Covariance described by Bollerslev and Wooldridge (1992).

2.2. Explanatory Variables

Our set of information variables, Χ

is derived from the literature on international stock market integration, which largely documents the explanatory power of some relevant economic variables in the determination of stock returns. These control variables are intended to predict the returns on the market indexes and to approximate the information that is available to investors when they set prices. The control variables used in our total information set are all lagged one period:

ƒ own excess returns,

ƒ exchange rate changes,

ƒ changes in the short term interest rates (30-day interest rates),

ƒ change in the long-short term spread (10-year government bond yield minus 3-6 months treasury bill yield),

ƒ and a Friday dummy to account for the week-end effect.

Studies related to international finance traditionally use two categories of predictable variables, a set of local components and a set of common or world components. Similarly, our information variables are divided into two categories. The first is a set of local instruments which are specific to each country. The second is a set of common instruments which are identical for all countries and which describes information from the EMU and the US. All these variables are designed to capture expectations about local and international economic conditions.

The first explanatory variable is the own excess returns. Several studies

such as Harvey (1991) include the lagged own-country return in the set of

instrumental variables to account for the degree of autocorrelation in stock

returns. The change in the exchange rates is the second explanatory

variable. Fratzscher (2001) uses this variable to explain the return on

European stock markets. The third variable is the change in the short

predictive ability on stock returns (Hjalmarsson, 2004). The change in the term structure spread is also included. The term structure spread is defined as the log difference in the yields on long-term government bonds and short-term Treasury bills. The long rate is the yield on 5-10-year government bonds. The short interest rates are measured as the yield on 3- month T-bills. The long-short yield has been used to predict excess U.S.

stock returns (Fama and French, 1989; Campbell and Hamao, 1992).

Several studies show that a term spread, in conjunction with one or more other variables, jointly predict returns on long-term corporate bonds and stocks. Fama and French (1989) conclude stock returns vary with a term spread. While most of the literature use a long-short term spread – the spread between a long-term yield and a short-term yield – to predict returns on long-term securities, Hjalmarsson (2004) finds that the term spread has strong predictive ability for stock returns in OECD countries only in the short-run. Finally, a Friday dummy is included to account for end-of-the-week effects and it is a common forecasting variable in studies of international stock returns. The predictability of our information variables is performed and discussed subsequently.

3 Data

Daily returns of stock indexes for the four Nordic countries (Denmark,

Finland, Norway, and Sweden) and for the EMU-area and the US stock

markets are analyzed. The returns are calculated as logarithmic differences

of the market indexes. The data set commences on February 1, 1996 and

ends on March 3, 2005, providing 2370 observations in total. All data is

obtained from Ecowin and is from Morgan Stanley. It is expressed in local

currency without dividend. We chose the Morgan Stanley indexes for

reason of harmonization in the data and as it is a widely used benchmark

for several studies in the literature. Mean, standard deviation, skewness,

kurtosis and autocorrelations of the daily log returns on the market indexes are provided in Table I.

Table I Summary Statistics for the Country Returns

R_DEN R_FIN R_NOR R_SWE R_EMU R_US

Mean 0.0404 0.0563 0.0299 0.0397 0.0281 0.0268 S.D. 1.1585 2.5914 1.2332 1.6929 1.3329 1.2323 Skewness -0.3192 -0.4125 -0.4348 0.0986 -0.2007 -0.0553

Kurtosis 5.4054 8.7205 6.6821 5.7969 5.5072 5.9807

ρ1 0.028

(0.174)

0.000 (0.999)

0.051 (0.013)

0.025 (0.230)

0.029 (0.164)

-0.026 (0.203)

ρ2 -0.046

(0.032)

-0.014 (0.799)

0.012 (0.040)

-0.008 (0.447)

-0.021 (0.225)

-0.023 (0.242)

ρ3 -0.021

(0.048)

-0.026 (0.567)

-0.031 (0.032)

-0.029 (0.315)

-0.049 (0.034)

-0.030 (0.176)

ρ4 0.020

(0.065)

0.006 (0.718)

0.040 (0.014)

-0.023 (0.313)

0.024 (0.039)

-0.014 (0.250)

ρ12 -0.055

(0.048)

0.018 (0.543)

0.003 (0.121)

0.024 (0.504)

-0.014 (0.005)

0.013 (0.151)

ρ36 -0.012

(0.039)

0.016 (0.171)

-0.011 (0.405)

-0.011 (0.005)

0.009 (0.000)

-0.009 (0.001)

The table reports the summary statistics for the daily returns (in %) on the market

indexes for Denmark (R

), Finland (R

), Norway (R

), Sweden (R

), the EMU (R

) and the US (R

). The statistics are based on daily data from 1996:2- 2005:3 (2370 observations). The country indexes are from Morgan Stanley, they are

expressed in local currency without dividends. ρ1, ρ2, ρ3, ρ4, ρ24, ρ36 express the first, second, third, fourth, twelfth and thirty-sixth autocorrelations of the series. The

p-values are provided in brackets.

All index returns exhibit excess kurtosis (kurtosis above 3) and the

skewness is negative for all except for Sweden. This reveals that the return

more frequent than positive shocks (negative skewedness) and that large shocks are more common than expected statistically (excess kurtosis).

Similar results are found in Fratzscher (2001).

The average daily return for the Nordic markets is between 0.0299%

(Norway) and 0.0563% (Finland). Sweden has the third mean return among the Nordic countries, after Finland and Denmark. The EMU and the US have almost similar average daily return (around 0.027%).

The standard deviation of the return series is quite dispersed. Finland has the highest standard deviation (2.59) and Denmark the lowest (1.16).

Sweden has a quite high standard deviation on the total panel of countries (the second highest after Finland). The standard deviations for the EMU and the USA are also quite analogous. On a portfolio investment perspective, given the choice between investing in one of the Nordic countries and the European and American portfolio, the risk-averse investor would be better off by investing in the Danish portfolio as it provides a fairly low standard deviation and a relatively high return. This is illustrated in Figure 3.

Figure 3 Portfolio Investment Choices

Return and Risk for the Country Indexes

Finland

Norway EMU US

Sweden Denmark

0 0,01 0,02 0,03 0,04 0,05 0,06

0 0,5 1 1,5 2 2,5 3

Standard Deviation

Mean

The daily average returns on the market indexes are plotted against their standard deviation. The average returns (in %) are from Morgan Stanley, expressed in local

currency without dividends. The sample period is from 1996:2-2005:3 (2370

observations).

The test statistics for serial correlation are quite divergent between the return series. While Finland is not at all plagued with autocorrelation, both Sweden and the US exhibit autocorrelation from the 36

lags. The EMU exhibits serial correlation at lags 1 and 2, Norway at the first fourth lags, and Denmark is plagued with autocorrelation form lag 2.

A starting point of many studies about financial integration is to look at the correlation coefficients among countries. It is common that correlation coefficients between national markets are modest in magnitude, except between countries which have closer links or in times when markets are affected by the same economic conjectures (see King et al., 1994). Table II provides the correlations between our return series. For our sample period, all correlations are positive and quite high.

The highest correlation coefficients are for the pairs Finland and Sweden, Finland and the EMU, and Sweden and the EMU. The lowest correlation is between the US and Denmark. As could be expected, the correlation between the EMU and the Nordic countries is higher than the correlation between the US and the Nordic countries.

Table II Correlation Coefficients of Daily Index Returns

R_DEN R_FIN R_NOR R_SWE R_EMU R_US

R_DEN 1 0.464 0.519 0.531 0.609 0.244

R_FIN 1 0.474 0.703 0.727 0.322

R_NOR 1 0.530 0.602 0.253

R_SWE 1 0.771 0.367

R_EMU 1 0.462

R_US 1

Table III reports summary statistics on the behavior of the common

information variables used in the study.

Table III Summary statistics for the Common Instrumental Variables

∆Eur/Usd ∆SR_EMU ∆TS_EMU ∆SR_US ∆TS _US Mean 2.56E-05 -0.000334 -0.000148 -0.000285 0.000393

S.D. 0.005942 0.010266 0.030657 0.014278 0.225408

Skewness 0.305019 -0.658327 1.136842 -1.412285 2.453051 Kurtosis 4.206405 27.81185 18.39823 34.20177 69.24297

ρ1 -0.005

(0.792)

-0.223 (0.000)

0.040 (0.054)

-0.195 (0.000)

0.112 (0.000)

ρ2 -0.015

(0.740)

0.066 (0.000)

0.017 (0.113)

0.064 (0.000)

-0.032 (0.000)

ρ3 -0.004

(0.887)

-0.019 (0.000)

-0.063 (0.003)

0.053 (0.000)

-0.151 (0.000)

ρ4 0.007

(0.944)

0.041 (0.000)

-0.031 (0.003)

0.034 (0.000)

-0.120 (0.000)

ρ12 0.025

(0.741)

0.009 (0.000)

0.007 (0.007)

0.000 (0.000)

0.062 (0.000)

ρ36 -0.012

(0.956)

0.021 (0.000)

-0.006 (0.006)

0.038 (0.000)

0.034 (0.000)

The statistics are based on daily data from 1996:2-2005:3 (2370 observations). The instrumental variables are: The change in the exchange rate between the Euro and the

US dollar (∆Eur/Usd), the change in the short-term interest rate expressed as the Eurocurrency rate for both the EMU and the US (∆SR

and ∆SR

), the change in

the term structure or long-short term spread expressed as the 10-year government bond yield minus 3-6 months treasury bill yield for both the EMU and the US (∆TS

and ∆TS

), a Friday dummy to cover end-of-the-week effects (Df).

The mean short-term change is negative both for the EMU and the US, indicating that the first lag tends to be lower than the one-period ahead.

The average term structure change for the EMU is also negative

illustrating that on average the Treasury bill yield is higher than the

government bond yield. Except for the change in Euro/Us dollar

exchange rate, all the control variable series are highly plagued with

4 Empirical Results

4.1. Predictability of Market Index Returns with the Explanatory Variables

In the empirical experiments, the set of information variables is split up into a local component part including country-specific variables, and a common component part which comprises an identical set of variables for all countries.

The local instrumental variables, all lagged one period, are own lagged returns, local exchange rate changes in relation to the Nordic currencies, the Euro, and the US dollar, changes in the local short term interest rates (30-day Eurocurrency interest rates), and the change in the local long-short term spread (10-year government bond yield minus 3-6 months treasury bill yield).

A similar set is considered for the common instruments: exchange rate changes between the Euro and the US dollar, short term interest rates changes (30-day Eurocurrency interest rates) for both the EMU and the US, the change in the long-short term spread (10-year government bond yield minus 3-6 months treasury bill yield) for both the EMU and the US, all are lagged one period. A Friday dummy is also included to account for the week-end effect.

The explanatory power of our variables differs across market indexes.

Overall, the most persistent variables are found to be the change in the exchange rates (for Denmark) between the euro and the Swedish krona, the Swedish krona and the US dollar, the US dollar and the Danish krona.

The change in the Swedish long-short term spread (for Finland and

Norway), and the 30-day American Eurocurrency rate (for Sweden) are

In order to capture the explanatory effect of the instrumental variables, we carry out a number battery tests. The starting point is to analyze the predictability of the returns on the four Nordic market indexes using the common set of instruments. Following Harvey (1991), we regress the returns in the market indexes on the common forecasting variables.

Estimation results are provided in Table IV. We report the p-values for significance of the parameters (in parentheses), as well as the R

statistic. It is found that the variables provide very little forecasting ability on the returns on the market indexes. Nevertheless, one variable has a significant effect for one of the market index returns (Sweden): The change in the short-term American interest rate. Before going to formal interpretation, we repeat the exercise this time by including the set of local information variables. The own-country variables as well as the country specific variables for the other Nordic countries are included in the regression.

Only the significant parameters are reported. Table V presents the results.

The R

statistics are given at the bottom of the table.

At the individual country level, the results are relatively divergent. The change in the SEK/USD exchange rate appears to be the sole variable with a significant effect for the return on the Danish market return. None of the variables seem to explain the return on the Finnish index returns.

For the Swedish market, two variables have explanatory power. First the change in the Finnish long-short term yield and second the Friday dummy.

The latter coefficient is, however, relatively small. Norway is quite

exceptional in the set of market indexes, as several variables have a

significant effect. The own lagged returns, the change in the EUR/USD

exchange rate, the change in the country specific long-short term yield, and

the change in the Finnish long-short term yield have explanatory power

for the market return series. Finally, the Friday dummy is found to predict

none of the return on the index series (except for Sweden). In summary,

table V provides considerable evidence that the returns on the Nordic

contrast, Campbell and Hamao (1992) find that common international factors generate expected returns in the US and Japan.

A comparison of the R

statistics provides an interesting result, i.e. the augmented specification adding country specific variables does no better in explaining returns than the starting model including only common variables. Indeed, the explanatory power is only marginally increased:

+0.20% for Denmark, +0.20% for Finland, +0.70% for Norway, and +0.40% for Sweden. This is strong evidence of the neutral effect of the information variables in explaining the returns for the Nordic markets.

The only local information variable that seems to be important for several countries is the Finnish long-short term yield for both Norway and Sweden. For comparison, in a study about the volatility patterns of the Swedish stock market using a threshold GARCH (TGARCH) and exponential GARCH (EGARCH), L. Berg, 2003 found that neither the exchange rates, nor the interest rates have a significant effect in the conditional mean equation. Using a GARCH(1,1), our results are similar for Sweden, and it seems that they could be extended to the other Nordic stock markets.

Interestingly, the intercept coefficient for each stock market index is found to be significant in all cases. A straightforward interpretation is that the coefficient might capture the explanatory power of omitted variables that we failed to include in our specification. Other studies on international stock returns have used such variables as the dividend-price ratios which we have arbitrarily excluded from our analysis.

4.2. Some evidence on Nordic Market Integration

Tables VI-1 to VI-8 report the estimates of financial integration for the

GARCH model (5)-(6) with constant coefficients. As specified in the

hand column provides the coefficients for the total sample period. The remaining right-hand columns show the coefficients for different sub- periods. Starting with an analysis for the total sample period, a number of striking results emerge.

A first look at the estimates reveals the dominance of the EMU over the US market for the four Nordic countries. This finding directly confirms our first hypothesis. Furthermore, it highlights not only the dominance of the EMU over the US market, but also its importance even for non-euro members. Our results are similar to Fratszcher (2001) who finds that, from 1986 through 2000, return spillovers from the Euro area market dominate return spillovers from the US for Denmark, Norway and Sweden. For our data, Finland is the most sensitive to return-spillovers from the EMU, and Denmark and Norway are the least. This intuitively relates to the EMU membership factor.

Finland and Sweden are sensitive to changes in each other’s stock markets.

Hence, a shock of 1% in the Swedish stock market triggers a change in returns of 0.0053% on average in the Finnish market and a 1% shock in Finland has a similar impact of 0,0031% on average in the returns on the Swedish market. Denmark and Norway are similarly joint receptive with an average change in returns of 0.0016% and 0.0018% respectively, caused by a 1% shock in either markets.

Comparing across countries for the importance of individual Nordic markets, Finland is more affected by shocks generated in Sweden than in any other Nordic stock market. Denmark (Norway) is more sensitive to shocks from Norway (Denmark) than to those from Sweden and Finland.

This is evidence that there is not a specific and exclusive leading market in the Nordic region.

Another interesting result is that Denmark and Norway are more

integrated with each other and with Sweden than with Finland. This

confirms the hypothesis that countries outside the EMU are more integrated with each others than with the participating country. However, Sweden appears to be most affected by return spillovers from Finland and thus more integrated with Finland than with Denmark and Norway. The close economic linkage between the two countries may partly explain this characteristic. Consequently, we fail to reject the second hypothesis as it is partly validated.

Finally, it is worth mentioning that own past shocks are not significant for any of the Nordic markets except Norway. This finding, coupled with the size of the coefficients for the shocks generated in the Nordic markets, is empirical evidence of how relatively integrated the four markets are over the whole sample period.

When looking at the coefficients for measuring volatility spillovers, the results pinpoint that the individual markets by far are most affected by own past volatility shocks. The size of the coefficients for the volatility shocks generated outside the national markets is small. However, some striking findings can be highlighted. Volatility in the Swedish stock market is far more increased by volatility shocks emerging from Finland, and the impact of those shocks generated in Denmark, Norway, the EMU and the US is somewhat analogous. Volatility shocks in Sweden appear to be the second source of volatility enhancement in the other Nordic markets after the shocks from the EMU. Finally, the coefficients suggest that the EMU may also be the dominant market for the Nordic countries in terms of volatility-spillover effects, as compared to the US.

4.3. Financial Integration over Time

As shown in Tables VI-1 to VI-8, different sub-periods are used in the

estimation. The whole sample period is split into ten sub-samples

are chosen so as to obtain equalized splits. The last sub-sample for the year 2005 is shorter as it only goes through early March.

This experiment enables us to extend the investigation one step further.

We can indeed observe the stability of the parameters through time and intuitively the changes over time of the degree of integration between the Nordic countries. Figure 4 plots the return-spillover coefficients over the ten sub-periods for each Nordic market.

A number of striking results are reflected. The first observation is that the coefficients of market integration before and after the event day have followed different patterns across the national markets. Overall, the integration of all Nordic financial markets with the EMU has substantially increased in the year which directly preceded the Euro introduction.

Formerly, it appears that the non-Euro members were not explicitly affected by the Euro advent. Only Finland exhibits a clear sign of increased integration from the start of the sample period. The sudden responsiveness of the non-Euro members at the dawn of the Euro launch is empirical evidence that the markets only reacted when the event day became formally confirmed and implemented.

Interestingly, all the Nordic countries display a decrease in integration with the EMU-area market after the euro introduction. This finding suggests that there is less evidence of price -or return-spillover effects form the EMU in the post-euro period. The integration with the US market appear to have remained constant over the sample period.

When looking across the Nordic markets, the integration between Sweden

and Finland has increased as time went towards the euro launch, and then

decreased substantially. Sweden and Denmark have become more

integrated up to the euro introduction and the pattern has stayed constant

in the post-euro period, but is reversed from the early 2004. While the

degree of integration between Sweden and Norway was decreasing

DENMARK

-0,005 0,000 0,005 0,010 0,015 0,020 0,025

2/96 -12/

96 1/97

-12/

97 1/98

-12/9 8 1/99-12/

99 1/00

-12/

00 1/01

-12/

01 1/02

-12/

02 1/03

-12/0 3 1/04-12/

04 1/05

-03/

05 Finland No rway Sweden EM U US

SWEDEN

0,000 0,005 0,010 0,015 0,020 0,025

Denmark Finland No rway EM U US

NORWAY

0,000 0,005 0,010 0,015 0,020 0,025

Denmark Finland Sweden EM U US

FINLAND

EM U

-0,005 0,000 0,005 0,010 0,015 0,020 0,025

2/96 -12/

96 1/97

-12/

97 1/98

-12/

98 1/99-12/

99 1/00

-12/

00 1/01

-12/

01 1/02

-12/

02 1/03-12/

03 1/04-12/

04 1/05

-03/

05 Denmark No rway Sweden EM U US

noticeably before the event day, it has increased constantly afterwards.

Except for a brief rise immediately followed by a drop between 2001 and 2003, the degree of integration between Denmark and Norway has kept quite constant over time, with no obvious impact from the euro introduction-day. This is straightforward since both countries do not participate in the EMU. However, a smooth increased integration between these two countries and Finland is observed just after the single currency has become effective.

The third hypothesis stresses that the Euro introduction has had a significant impact in the integration of the Nordic markets. The results provided here allow us to reject this hypothesis, as it is shown that the Euro launch has only a trivial weight in the integration of the Nordic countries.

Figure 4 Return Spillovers for the Nordic Stock Markets

5 Conclusion

There is increasing evidence that the completion of the European Monetary Union, through the introduction of the Euro, has built up the linkages between the capital markets of the participating countries.

However, the attention to the financial integration in Europe has been concentrated to the participating countries.

In this study we analyze the integration of the four Nordic financial markets from the early 1996 through the early 2005. We address three hypotheses. The first hypothesis states that the EMU-area market is the dominant market for the Nordic countries. We fail to reject this hypothesis as the dominance of the EMU over the US in explaining returns on the Nordic countries is significantly confirmed by our results.

The second hypothesis stresses that the non-EMU members should be more integrated with each other than with the Finnish participating country. Correspondingly, a higher and exclusive integration of Finland with the EMU than with any of the Nordic markets is expected. Empirical evidence underlines that only Denmark and Norway are more integrated with each other. Sweden is found to be more integrated with Finland than with Denmark and Norway. The second hypothesis can therefore be only partly rejected.

In the third hypothesis, we expect that the Euro introduction has had a

significant impact on the integration of all the Nordic markets. There are

clear signs that the degree of integration between Nordic countries has

changed through time, and that this change over time displays different

patterns across countries. The degree of integration for the pairs between

Finland and the other Nordic countries seems clearly related to the

introduction of the Euro. The event day, on the other hand, appears quite

neutral for the pairs between non-Euro members. A uniform pattern

integration of the Nordic countries with the EMU-market after the Euro launch. Intuitively, these findings suggest that financial market integration in the Nordic region has not been significantly influenced by the European unification process and we can therefore reject the third hypothesis.

A particular feature of financial integration is that it tends to be

asymmetric – negative shocks are more strongly spread than positive

shocks – and that they are often skewed –large shocks tend to have larger

effects than small shocks. The GARCH framework used in this study has

not accounted for asymmetric effects in shock transmission. Therefore,

further research could be conducted on the nature of asymmetric shocks

across the Nordic financial markets, by using more complex forms of

GARCH models.

References

Adler, M., and B. Dumas, 1983. International Portfolio Choice and Corporation Finance: A Synthesis, Journal of Finance 38, 925-984.

Bekaert, G., and C.R. Harvey, 1995. Time-Varying World Market Integration, Journal of Finance volume 50, 103-444.

Berg, L., 2003. Deterministic seasonal volatility in a small and integrated stock market: the case of Sweden, Finnish Economic Papers 16, number 2, 61-71.

Bessler, D. A., and J. Yang, 2003. The Structure of Interdependence in International Stock Markets, Journal of International Money and Finance 22, 261-287.

Bollerslev T., 1986. Generalized Autoregressive Conditional Heteroskedasticity, Journal of Econometrics 37, 307-327.

Bollerslev, T., and J.M. Wooldridge, 1992. Quasimaximum Likelihood Estimation and Inference in Dynamic Models with Time-Varying Covariances, Econometric Reviews 11, 143-173.

Campbell, J.Y., and Y. Hamao, 1992. Predictable Stock Returns in the United States and Japan: A study of Long-Term Capital Market Integration, Journal of Finance 47, 43-69.

Dumas, B. and B. Solnik, 1995. The World Price of Foreign Exchange Risk, Journal of Finance 50, 445-479.

Engle, R.F., 1982. Autoregressive Conditional Heteroscedasticity with Estimates of the Variance of UK Inflation, Econometrica 50, 987–1008.

Errunza V., E. Losq, and P. Padmanabhan, 1992. Tests of Integration, Mild Segmentation and Segmentation Hypotheses, Journal of Banking and Finance 16, 949–972.

Errunza, V. R., and E. Losq, 1985. International Asset Pricing under Mild

Segmentation: Theory and Test., Journal of Finance 40, 105-124.

Fama, E.F., and K.R. French, 1989. Business Conditions and Expected Returns on Stocks and Bonds, Journal of Financial Economics 25, 23-49.

Ferson, W.E., and C.R. Harvey, 1993. The Risk and Predictability of International Equity Returns, Review of Financial Studies 6, 527-566.

Frankel, J.A., 1994. The Internationalization of Equity Markets, NBER Working Papers, 1-36.

Fratzscher, M., 2001. Financial Market Integration in Europe: On the Effect of EMU on Stock Markets, Working Paper No.48, European Central Bank.

Harvey C.R., 1991. The World Price of Covariance Risk, Journal of Finance 46, 111-157.

Hjalmarsson, R., 2004. On the Predictability of Global Stock Returns, Working Papers in Economics No. 161, School of Economics and Commercial Law, Göteborg University.

King M., E. Sentana, and S. Wadhwani, 1994. Volatility Links between National Stock Markets, Econometrica 62, 901-933.

Longin, F., and B. Solnik, 1995. Is correlation in International Equity Returns Constant: 1960-1990, Journal of International Money and Finance 1995, 3-26.

Solnik, B. H., 1983. International Arbitrage Pricing Theory, Journal of Finance 38, 449-457.

Stulz R.M., 1981. On the Effects of Barriers to International Investment, Journal of Finance 36, 923-934.

Wheatley S., 1988. Some Tests of International Equity Integration, Journal of Financial Economics 21, 177-212.

Yang J., J.W. Kolari, and G. Zhou, 2004. Real Estate Market Integration in

Europe: Evidence from the Stock Market.

Tables

Table IV

Forecasting Market Index Returns with Common Instrumental Variables

µ β_Eur/USD δ_i,EMU δ_i,US φ _i,EMU φ _i,US Df R²

Denmark 0.00085

(0.00030)

-0.05177

(0.1706)

0.01826

(0.3279)

0.00751

(0.5670)

-0.00101

(0.8845)

-0.00113

(0.3136)

0.00041 (0.3527)

0.366 Finland 0.00090

(0.0073)

-0.00272

(0.9606)

-0.01484

(0.4286)

0.03058

(0.1134)

0.00243 (0.8505)

0.001050 (0.6238)

0.00102 (0.1371)

0.507 Norway 0.00075

(0.0002)

0.03701

(0.2183)

0.01764

(0.3508)

0.00526

(0.6845)

0.00507

(0.4840)

0.00192

(0.1420)

0.00054 (0.2150)

0.352 Sweden 0.00106

(0.0000)

0.00088 (0.9286)

0.00510 (0.3959)

0.00960 (0.0271)

0.00628 (0.0583)

-7.17E-05 (0.9335)

-0.00012 (0.3666)

0.555

The regressions are based on daily data from 1996:2-2005:3 (2370 observations). The common set of instrumental variables is composed of:

exchange rate changes between the Euro and the US dollar (∆Eur/USD), short term interest rates changes (30-day Eurocurrency interest rates) for the EMU (∆SR_EMU) and the US (∆SR_US), the change in the long-short term spread (10-year government bond yield minus 3-6 months treasury bill yield) for the EMU (∆TS_EMU) and the US (∆TS_US), and a Friday dummy to account for the week-end effect (Df). The control variable series are calculated as the first logarithmic differences and they are all lagged one period (except the Friday dummy). The GARCH(1,1) specification for the regression on each individual country index is estimated as:

1 , , , ,

1 , 1 , ,

1 , ,

1 , ,

1 , ,

1

, / ₋

=

−

−

−

−

− + ∆ + ∆ + ∆ + ∆ + + +

∆ +

=

∑

i j j i t

US US i t EMU EMU

i t US US i t EMU EMU

i t i

t

i Eur USD SR SR TS TS

R µ β δ δ φ φ λ ε λ ε λ ε (9)

2 1 , , 2

, ,

2 , 1

, 2

1 , 2

1 , 2

, −

=

−

− + + + +

+

=

∑

i j i t

i i t i i i t

i ω βσ αε α ε α ε α ε

σ (10)

The idiosyncratic shocks for the respective markets are included in the specification and computed as stated in equations (3)-(4). The market indexes are from Morgan Stanley, provided in local currency without dividends. The p-values are given in brackets. Concerning the p-values, no attempt has been done in order to correct these due to the fact that predicted variables are included in the estimation.

Table V

Forecasting Market Index Returns with Common and Country Specific Instrumental Variables

Denmark Finland Norway Sweden

µ 0.000864

(0.0000) 0.001091

(0.0010) 0.000886

(0.0000) 0.000690

(0.0007)

α 0.026711

(0.2273) 0.007501

(0.7640) 0.056574

(0.0112) -0.003629 (0.8756) Df -0.000345

(0.3384) -0.000406

(0.5980) -0.000107

(0.8018) 0.000960 (0.0293)

β1 βEUR/DKK 0.161060 (0.6692) - βEUR/NOK 0.023991 (0.9264) βEUR/SEK 0.343617 (0.1124)

β2 βUSD/DKK -0.346427 (0.1135) - βUSD/NOK 0.067086 (0.7980) βUSD/SEK -0.061379 (0.7947)

β_j/i β_SEK/DKK - 0.058925 (0.3464)

βNOK/DKK -0.058925 (0.3461)

β_EUR/SEK 0.204426 (0.4072) βEUR/DKK 0.844755 (0.1766) βEUR/NOK -0.360902 (0.4591)

β_SEK/NOK 0.023992 (0.9264) βDKK/NOK 0.032798 (0.6024)

βDKK/SEK -0.040349 (0.6200) βNOK/SEK 0.473474 (0.0872)

The regressions are based on daily data from 1996:2-2005:3 (2370 observations). The local set of instrumental variables is composed of: The own lagged retun (R_i,t-1); the change in the exchange rate between the local currency and the Euro, the US dollar, and the other Nordic currencies (respectively ∆Eur/i, ∆Usd/i, ∆j/i with i=individual country and j=all other Nordic countries); the change in the local short-term interest rate (∆SR_i) and the change in the local or long-short term spread expressed as the 10-year government bond yield minus 3-6 months treasury bill yield (∆TS_i). The common set of instrumental variables is composed of: Exchange rate changes between the Euro and the US dollar (∆Eur/USD), short term interest rates changes (30-day Eurocurrency interest rates) for the EMU (∆SR_EMU) and the US (∆SR_US), the change in the long-short term spread (10-year government bond yield minus 3-6 months treasury bill yield) for the EMU (∆TS_EMU) and the US (∆TS_US), and a Friday dummy to account for the week-end effect (Df). The sum of the common variables are expressed as a vector of variables X’ with k=EMU, US. Only the common instruments with a significant effect are reported in the table. The control variable series are calculated as the first logarithmic differences and they are all lagged one period (except the Friday dummy). The GARCH(1,1)