Interdependencies of Markets in Southeastern Europe and Buyback of Shares on Shallow Capital Markets: The Application of Cointegration and Causality Tests

Acknowledgement: This paper is a part of research project No. 179065, 47009 and 179015 supported by the Ministry of Education and Science of the Republic of Serbia. The authors gratefully acknowledge two anonymous referees whose comments improved this article. Interdependencies of Markets in Southeastern Europe and Buyback of Shares on Shallow Capital Markets: The Application of Cointegration and Causality Tests

In the environment of drawn out effects of global financial crises (GFC), stock exchange indices in the countries of Southeast Europe (SEE) are either stagnating or slowly recovering.Moreover, in the case of issuers with good business results and with actively traded shares, share prices may fall faster than the representative stock exchange index during a certain period of time.
By applying the cointegration tests and Granger causality test to monthly data series, in the period from 2007 to 2011 and by assessing the equilibrium/error correction model, this paper examines: firstly, interdependence of stock exchange indices' movements in the SEE countries (Slovenia, Croatia, Serbia, Montenegro, Republic of Srpska, Macedonia -FYRM, Bulgaria), with a focus on the interdependence of movements of indices in Serbia -Belex15 index and stock exchange indices in the neighbouring countries; secondly, interdependence of the movements of the most traded shares on the capital market in Serbia -AIK Banka (AIK) shares and the movements of Belex15 index.AIK is among the ten largest banks in Serbia.In the PANOECONOMICUS, 2015, Vol.62, Issue 4, pp.469-491 period from 2007 to 2010, it ranked either first or second based on several key performance indicators -return on capital, return on equity, share of operating expenses in operating revenue, and the capital adequacy ratio.An important characteristic of AIK shares is that they were most heavily traded on the Belgrade Stock Exchange.Despite good business results of AIK, the price of its shares recorded a sharp drop after the onset of GFC which spilled over to the capital market in Serbia in 2008.
Considering the above, the question arises whether the management of AIK can use its good business results that were effectively achieved to influence the behaviour of the AIK share prices on the shallow domestic capital market while taking into account the negative effects of external factors relating to the protracted effect of the GFC.A question arising from the above is whether it is possible to effectively use the share buyback as a tool to stop or mitigate negative trends of the AIK share price.

Literature Review
Markets in SEE before and after the emergence of the GFC have not been elaborated in the recent literature by applying the quantitative methods of interdependences.The use of buyback of shares on shallow capital markets to defend the share price of banks with good business performance when the management believed that share price did not accurately reflect the bank's prospects has not been analysed.The recent literature mostly considers the movement of share prices of corporations and banks on developed markets and buyback of shares before and after the emergence of the GFC for the purpose of not only defending the share price, but also of optimising the bank's capital structure (Paul Brockman, John S. Howe, and Sandra Mortal 2008; Chun I. Lee, Demissew Diro Ejara, and Kimberly C. Gleason 2010;Amedeo De Cesari, Susanne Espenlaub, and Arif Khurshed 2011;Dev Mishra, Marie D. Racine, and Luke Schmidt 2011;Jonas Råsbrant 2011;Steven Mintz 2012;The Economist 2012;In-Mu Haw et al. 2013).
Authors who examined the share price effects and determinants of share repurchase programs of companies in developed market economies (France, Germany, Italy, and UK) before the GFC, indicated that share prices reaction to share repurchase could be both positive and negative.German and Italian share repurchases were followed by a positive and significant share price response (similar to US firms).In the case of UK they were followed by small positive abnormal returns, and in the case of France, there were abnormal returns which were insignificantly different from zero (Lee, Ejara, and Gleason 2010, p. 125).Konan Chan, David L. Ikenberry, and Inmoo Lee (2007), also reported evidence of abnormal stock performance following buyback announcements.Based on an extensive examination of 5508 programs announced by US firms in the period from 1980 to 1996, their findings indicated abnormal stock performance following buyback announcements.However, when examining the factors which led to the observed share price reactions following the buyback announcement, Chan, Ikenberry, and Lee (2007) concluded that the movement of prices was more consistent with the view that "managers possess market timing abilities when announcing and executing buyback decisions" and that pseudo-market timing would not be the viable explanation (p.2692).Using a sample PANOECONOMICUS, 2015, Vol.62, Issue 4, pp.469-491 that corrects problems of samplings in previous literature and the calendar portfolio regression method, Ken C. Yook (2010) finds that "firms that announce repurchase programs infrequently and repurchase shares actually experience significant longterm abnormal returns" (p.323).On the other hand, Sheng-Syan Chen and Yanzhi Wang (2012) find that financial constraints of repurchasing firms especially those with high actual repurchase ratios, "experience a significantly greater increase in post-buyback distress risk than unconstrained firms.Managerial hubris could explain why constrained firms buy back shares even if the buybacks do not improve shareholder wealth" (p.311).Cristiana Tudor (2011) investigated interdependencies among six selected stock exchanges in the CEE region (Czech Republic, Hungary, Bulgaria, Poland, Russia and Romania) from 2007 to 2009 and concluded that the interdependencies among CEE stock markets were generally larger during the crises than before as well as that the overall dominant role and the impact of the USA market on the CEE exchanges seemed to have decreased during the crises.
During the GFC many companies in developed market economies used share buyback (Leonce Bargeron, Manoj Kulchania, and Shawn Thomas 2011;Damien Farrugia, Michael Graham, and  Fortuna Advisors, a New York City-based consultancy, prepared the Institutional Investor's 2012 Corporate Buyback Scorecard that ranked the biggest spenders on buybacks among the S&P 500 based on the two-year returns generated by their repurchases.Institutional Investor's 2012 Corporate Buyback Scorecard which referred to 253 companies that bought back at least 4 per cent of their market capitalization in a two-year period (Q2 2010 -Q2 2012), indicated that more than a half of the 253 companies recorded negative buyback effectiveness (return on investment) "indicating that managers generally time buybacks poorly" (Mintz 2012, p. 2).During the same period Goldman Sachs surprisingly realised a USD 9.8 billion shares buyback program with 20% negative return on investment.Kathleen Farrell, Emre Unlu, and Jin Yu (2014) find that the use of share repurchases as an earnings management mechanism is discouraged in the presence of debt-financing constraints.
Certain banks from developing markets in Europe announced their buyback of shares programmes during the GFC period.In SEE, there were two cases -in Hungary (the case of OTP Bank, Hungary's largest retail bank with the largest asset sum) and in the Czech Republic (Komercni Banka AS).In the case of OTP Bank, in May 2010, shareholders adopted a plan based on which the Board was allowed to buy back shares provided the number of shares that OTP Bank kept in its own treasury did not exceed 56 million shares or 20 per cent of the bank's registered equity as of October 31, 2011.According to the Global Banking News (2010), the Board was allowed to buy the shares at a price not higher than 120 per cent of the closing price of OTP shares on the Budapest Stock Exchange by October 31, 2011.At the end of June 2010 the average price of shares was HUF 4.626, while in the first ten days of January 2011, when the share buyback program was implemented, the average price was HUF 5.205.At the end of May 2011, the price was HUF 6.303.From June to October 2011, OTP share prices decreased and rebounded, and the portfolio of bank's own shares jumped from 1.49% to 1.66% at the end of October 2011.Due to the increase in price in 2012, selling of shares that were bought back would have increased the profit.Based on the share price movement and total amount of shares held by the bank, one is to conclude that OTP Bank's share buyback programme was successfully implemented.As for Komercni Banka AS, due to the lack of available data, it is not possible to draw the reliable conclusions with regard to the success of the buyback programme implementation.

Empirical Method
The analysis of interdependences requires the examination of their stationarity by applying the unit root test.If the time series are stationary, the regression analysis can be used.If the time series are not stationary it is necessary to verify if the cointegration exists.If the cointegration exists, the next steps would be to implement the analysis of causality and error correction model.
We examine the presence of causality in the Clive W. J. Granger (1969) sense at the following two planes: first, there is an examination of the correlation between stock exchange indices' movements in the SEE countries; second, there is an examination of the extent to which the behaviour of the prices of the most traded shares on the Belgrade Stock Exchange (AIK shares), affects the behaviour of the prices of all shares on the Belgrade Stock Exchange, measured by the representative index -Belex 15.This is the index in which AIK shares have the largest weight of 20%.

Data
The analysis covers a 5-year period, from January 2007 to December 2011.The monthly data series are long enough to allow identifying the patterns in the examined events and drawing reliable conclusions.Key time sequences in the analysed period are as follows: the GFC broke out after the subprime mortgage market in the U.S. collapsed in August 2007.The GFC affected SEE countries in 2008, most intensely in Q4 of 2008.

Synchronized Movements of Stock Exchange Indices in SEE Countries
The graph of stock exchange indices in the analysed period from January 2007 to the end of 2011 shows that similar trends are present in all stock exchange indices, with the exception of Montenegro stock exchange index -Monex20 where movements are characterised by an exceptionally high degree of variability.
The analysis of correlation matrix shows that there is a high positive correlation between all stock exchange indices (see their full names in Appendix A), with Belex15 and Crobex standing out with the highest correlation coefficient of 0.9658.Setting apart of Monex20 (index) from other indices is confirmed, in addition to the PANOECONOMICUS, 2015, Vol.62, Issue 4, pp.469-491 graphical presentation, by the correlation analysis, with the lowest correlation matrix coefficients being present between the Monex20 and other indices (Table 1).Due to the weakness of correlation analysis, such as a possible presence of false correlation in the case of non-stationary time series, it is followed by the stationarity analysis.

Cointegration Analysis
An exceptionally high level of positive linear correlation between the observed series suggests that a movement synchronisation between relevant stock exchange indices may be present.The first step of the analysis requires checking the observed time series stationarity.Testing is performed at the significance level (error in conclusion) of 0.05, with the null hypothesis rejected if the resulting significance level (p-value) is less than 0.05.In the example of all observed stock exchange indices (Table 2), the null hypothesis is not rejected and in the first phase of the test we conclude that every time series has at least one unit root.In the second phase of testing, we reject the null hypothesis at about at least two unit roots for every time series and conclude that every time series is non-stationary.
Based on Table 2, all time series have precisely one unit root, which allows for the analysis of the presence of cointegration.An equal degree of non-stationarity (equal number of unit roots) is a necessary precondition for testing the presence of time series cointegration.For more details, see: Russell Davidson and James G. MacKinnon (1993), James D. Hamilton (1994), Fumio Hayashi (2000), Helmut Lütkepohl (2005), Katarina Juselius (2006), Milena Jovičić andRadmila Dragutinović Mitrović (2011).
Since in the cointegration relation any time series may appear in the role of either a dependent or independent variable, we will have two cointegration tests for each time series pair (Table 3).Since the rejection of the null hypothesis implies the stationarity of the residual of observed cointegration time series combination, we conclude that only the stock exchange indices of the Republic of Srpska and the Republic of Serbia are cointegrated.Other cointegration relationships do not exhibit the stationarity of accidental error.The non-stationarity of the cointegration relation stochastic variable is, in most of the analysed time series, a consequence of the presence of structural breaks.The above conclusion is apparent based on the kurtosis coefficient analysis of each time series, but also on the auxiliary regression residual series in testing the non-stationarity.Each time series characterised by a non-standard movement trend will affect the kurtosis of the given series which, in most cases, will be greater than in the normal distribution.Kurtosis coefficients are manifestly different in every analysed time series because of different sharpness, depth, and durability of structural breaks, but also because of the point of time in which they occur.
Consequently, it can be concluded that the flows of analysed time series are not equally affected by the global financial downturn.Stock exchange indices Be-lex15 and BIRS are interesting examples of the presence of cointegration.These indices are also characterised by the presence of structural breaks, which is reflected in the residuals of cointegration relation auxiliary regression in which the dependent variable is the time series differential.Structural breaks are manifested in the form of two non-standard observations in May 2008 and October of the same year, simply modelled by artificial variables, after which valid conclusions about the cointegration of observed time series around the given structural breaks can be made.In other time series, it is not possible to model the structural breaks with artificial variables, due to the differences in structural breaks in those time series.

Causality Analysis
The existence of cointegration between the observed time series allows for the application of the analysis of the direction of causality between the observed time series.The first phase of the analysis involves the evaluation of the regression equation in which the dependent variable is presented depending on its own earlier values from one, two, or more periods, but also the earlier values of other time series that may be statistically significant and may confirm the direction of causality.However, if an equivalent conclusion is made in the case of other analysed time series, we conclude that the time series movements are synchronised, but that unidirectional causality is not present.
Based on Table 4, we conclude that there is an influence of the BELEX15 stock exchange index movements on the BIRS stock exchange index movements.The above mentioned structural breaks are modelled with the artificial variables V1 and V2.No residual autocorrelation or disturbance of the assumption about the normality of distribution is present in the evaluated regression model.The null hypothesis about the normality of distribution is not rejected at the level of significance of 5% regardless of the low achieved level of significance resulting from the higher kurtosis coefficient.According to Juselius (2006), a kurtosis coefficient that is higher than normal is often ignored.Table 5 suggests that there is a lower degree of inertia in the movements of BELEX15 series since, among the independent variables in the model, its own trend from only one earlier period is present.On the other hand, considering the dependence on their own earlier values from three previous periods, BIRS time series reflect a considerably higher degree of inertia in their movements.In the BELEX15 model, among the dependent variables there are 4 artificial variables, two of which are related to the structural break discovered in the unit root test, and the other two relate to April and May 2007.The model does not have any residual autocorrelation PANOECONOMICUS, 2015, Vol.62, Issue 4, pp.469-491 up to the 12 th lag (the 12 th lag is the one taken into account because of data frequency -monthly).Normality test also shows that the assumption about the normality of distribution is satisfied (p-value=0.44).In the error correction model (ECM) for BELEX15, the dependent variable is presented in the form of a differential (its stationary presentation) in the function of an equilibrium error (the stochastic variable of cointegration relation).Based on the evaluated coefficient, we conclude that 20% of the change in BELEX15 index is, at any point in time, adjusted to the common movements it forms with the BIRS index.Evaluation of the BIRS index in the model is not statistically significant and therefore, will not be interpreted.According to Zorica Mladenović and Aleksandra Nojković ( 2008), in the case of cointegrated time series, at least one must be presented in the form of the ECM.The first sub-period, in which the given time series exhibited growth, ended in the middle of QII 2007, namely in May, when their maximum values were reached.The second sub-period lasted until the beginning of QII 2009, in which the BE-LEX15 index and AIK share prices were falling, with the sharpest drop recorded in QIV 2008.QI 2009 is non-standard due to the greatest difference in the series movements achieved in January.At that time, the price of AIK shares grew by 30% compared with the price in December 2008, while the BELEX15 index remained the same.

Synchronized Movements of the
During the third sub-period, the series movements stabilised, but with two specific characteristics: the first relates to a more pronounced rise of AIK share prices than that of the BELEX15 index from the beginning of QIV 2010 to January 2011, based on the difference in the monthly price rise rates (in the given sub-period it peaked out in February 2011), after which the price of AIK shares declined faster.After January 2011, the greatest difference in series monthly growth rates in favour of the BELEX15 index was recorded in June 2011, when the monthly growth rate of AIK share prices was by 17 percentage points below the monthly growth rate of the BELEX15 index, most likely owing to the rumours about AIK's exposure to NI-BENS Grupa.By releasing information that it set aside provisions for loans extended to NIBENS Grupa (EUR 40 million), AIK put to rest the rumours about its excessive exposure to losses against the loans granted to this Group through Bloomberg Company in direct communication with largest investors and through domestic media.

Cointegration Analysis
The unit root test (time series stationarity) examines the alternative hypothesis about the stationarity of time series, as opposed to the null hypothesis about having at least one unit root.The testing ends with the rejection of the null hypothesis and the determination of the number of the time series unit root (Table 7).The ADF test is performed at the significance level of 5% (conclusion error) and the null hypothesis is rejected in the second testing phase (the realised level of significance, namely p-value, is less than 5%) and it is concluded that both time series have precisely one unit root, namely that they have the same order of integration.The existence of the same order of integration in the analysed time series enables us to test the presence of cointegration (long-term synchronization of movements) for the given period.
The cointegration test is a unit root test on the residuals of cointegration relation (in the regression of the BELEX15 index on the prices of AIK shares, and the other way round).In the first step, the regressions of one time series to the other are evaluated, and then the unit root test is performed on the model residuals, as indicated in Table 8.In the cointegration relation (regression model), where dependent variable is the BELEX15 index, we conclude, using the time series cointegration test, that relevant time series are cointegrated.In the case where dependent variable is AIK share price, we conclude, using the cointegration test, that the observed time series are not cointegrated.The obtained result suggests that there exists a unidirectional causality between AIK share prices and the BELEX15 index, which is further confirmed by the Granger causality test.

 Causality Analysis between the BELEX15 Stock Exchange Index Movements and AIK Share Price Movements
The Granger causality test is based on the regression of a dependent variable to its own values from the previous periods and to the values from earlier periods of other independent variable.It is necessary to evaluate the model in which the assumptions about stochasticity are satisfied.The Granger causality test determined that there was no unidirectional influence of the BELEX15 index on the movements of the prices of AIK shares, which was to be expected.No autocorrelation or an impaired assumption about the normality of residual distribution is present in the evaluated model.Considering that they are cointegrated, time series can be presented in the form of an error correction model.

Error Correction Model
In the case of time series cointegration, it is possible to evaluate the ECM.The ECM is based on the regression of stationary transformation of the dependent variable to the cointegration relation stochastic error (equilibrium error) from the previous period.The coefficient with the equilibrium error from the previous period shows which part of the change of dependent variable is, in each period, adjusted to the evaluated equilibrium relation.The main advantage of the ECM is that it includes in the analysis the information about time series at the level, rather than their transformations.Time series at the level are included in the ECM since they are elements of the equilibrium error.In the ECM for AIK residuals, cointegration relations with a time-lag of one period are not statistically significant, which is confirmed by the direction of causality between the given time series (Table 11).In the ECM for BELEX15, residuals of cointegration analysis are statistically significant, which confirms the direction of causality between the given series (Table 12).The model satisfies the specification tests, namely the assumptions about the accidental element.The coefficient with the equilibrium error shows the adjustment of BELEX15 index, at any point in time, to a long-term equilibrium relation it creates with AIK share prices.It can be concluded that 28.15% of the BELEX15 index change, at any point in time, is adjusted to the path of the long-term equilibrium relation between the relevant index and the AIK share price movements.To put it differently: almost 30% of the absolute change of the value of BELEX15 index in every month compared with the previous one is adjusted to the common long-term path created by BELEX15 and AIK share prices (since the evaluated coefficient with the equilibrium error from the previous period, representing a deviation from the equilibrium path, equals 0.2815).Artificial variables covering the presence of reviewed structural breaks and the first differential of the prices of AIK shares, modelling the short-term dynamics that is not relevant for interpretation in the ECM, are present in the model.

Analysis of the Monthly Growth Rates of AIK Share Price and BELEX15 Index
Source: Authors.6).The average annual drop of BELEX15 in the observed period was -9.46%.
A comparative analysis of marked drops and surges in AIK share prices and BELEX15 index shows that there is a greater drop in AIK share prices in the periods of marked decline of both series, and a greater rise in AIK share prices in the periods of marked growth of both series.In March 2007 and January 2009, the months of significant growth in AIK share price, a marked rise in the BELEX15 index monthly rate cannot be observed.However, statistical significance of the differences between the monthly growth rates for the observed series may be analysed by the time series stationarity of the difference between monthly growth rates of AIK share prices and BELEX15 index.
Source: Authors.Relevant "positive" breaks for AIK are two out of five "positive" breaks, that is, a surge in AIK share prices (when the monthly growth rate exceeds 30%) as shown in Figure 7.The negative break of the difference between the monthly growth rates is one out of two negative breaks of the monthly growth rates AIK share prices.A statistical significance of these breaks from Figure 7 is confirmed in the regression model in Table 13 where the breaks are covered by artificial variables VV1, VV2, and VV3.It can be concluded that in the case a monthly growth rate of AIK share prices is above 30%, the BELEX15 index may have a considerably higher monthly growth rates, and vice versa, while in the case of lower rates of AIK share prices, the differences in monthly growth rates are of stochastic nature, which is confirmed by the stationarity test of the given time series as indicated in Table 14.small number of banks from SEE which were able to use share repurchase before and during the GFK.Prerequisites for that are a multiyear active trading in bank shares on stock exchanges and above average bank (business) performance.

Figure 2
Figure 2 Autocorrelation Test and Test of the Normality of Distribution

Figure 3
Figure 3 Autocorrelation Test and Test of the Normality of Distribution Belex15 Stock Exchange Index and Price of the Most Traded Share (AIK Share Price) Source: Belgrade Stock Exchange (2013).

Figure 4
Figure 4 BELEX15 Index and AIK Share Movements A visual analysis of the BELEX15 index and AIK share price movement trends reveals that the analysed time period from January 2007 to December 2011 has three different sub-periods.During the same period one should note that according to the 12 business performance indicators, AIK was among the top 10 banks (see Appendix B).The first sub-period, in which the given time series exhibited growth, ended in the middle of QII 2007, namely in May, when their maximum values were reached.The second sub-period lasted until the beginning of QII 2009, in which the BE-LEX15 index and AIK share prices were falling, with the sharpest drop recorded in QIV 2008.QI 2009 is non-standard due to the greatest difference in the series movements achieved in January.At that time, the price of AIK shares grew by 30% compared with the price in December 2008, while the BELEX15 index remained the same.During the third sub-period, the series movements stabilised, but with two specific characteristics: the first relates to a more pronounced rise of AIK share prices than that of the BELEX15 index from the beginning of QIV 2010 to January 2011, based on the difference in the monthly price rise rates (in the given sub-period it peaked out in February 2011), after which the price of AIK shares declined faster.After January 2011, the greatest difference in series monthly growth rates in favour of the BELEX15 index was recorded in June 2011, when the monthly growth rate of AIK share prices was by 17 percentage points below the monthly growth rate of the BELEX15 index, most likely owing to the rumours about AIK's exposure to NI-BENS Grupa.By releasing information that it set aside provisions for loans extended to NIBENS Grupa (EUR 40 million), AIK put to rest the rumours about its excessive The movements of BELEX15 index depend on the movements of its own value from the previous month and the movements of AIK share prices from the previous two months.The artificial variables comprising the structural breaks (sub-periods) in the movements of relevant time series explained in the descriptive analysis, are present in the model.No artificial variables are present in the model of the AIK share price movements as the distribution of stochastic error does not deviate from normal.Granger causality test confirmed a high influence of the movements of AIK share prices on the BELEX15 index (significance of the influence of the movements of AIK share prices with a two-month lag).The residuals of this model satisfy the specification tests (the absence of autocorrelation and impairment of the assumption about normality).

Figure 5
Figure 5 Monthly Growth Rate of AIK Share Price, in % The analysis of the graphical presentation of monthly growth rates of AIK share prices (Figure 5) shows that they surged in March and April 2007, January 2009, May 2009, and September 2009, as marked with grey circles below.A sharp drop in AIK share price was recorded in October 2008 and March 2009, as marked with blank circles.The average annual drop in AIK share price in the observed period is -9.55%.

Figure 6
Figure 6 Monthly Growth Rate BELEX15 Index, in % The BELEX15 index monthly growth rate peaked in April 2007, May 2009, and September 2009.A considerable drop of the given series was seen in October 2008 and March 2009 (Figure6).The average annual drop of BELEX15 in the observed period was -9.46%.A comparative analysis of marked drops and surges in AIK share prices and BELEX15 index shows that there is a greater drop in AIK share prices in the periods of marked decline of both series, and a greater rise in AIK share prices in the periods

Figure 7
Figure 7 Difference between the Monthly Growth Rates of AIK Share Prices and BELEX15 Index Figure 7 suggests that two "positive" structural breaks for AIK took place in March 2007 and January 2009, while one negative structural break occurred in March 2009.Relevant "positive" breaks for AIK are two out of five "positive" breaks, that is, a surge in AIK share prices (when the monthly growth rate exceeds 30%) as shown in Figure7.The negative break of the difference between the monthly growth rates is one out of two negative breaks of the monthly growth rates AIK share prices.A statistical significance of these breaks from Figure7is confirmed in the regression model in Table13where the breaks are covered by artificial variables VV1, VV2, and VV3.It can be concluded that in the case a monthly growth rate of AIK share prices is above 30%, the BELEX15 index may have a considerably higher monthly growth rates, and vice versa, while in the case of lower rates of AIK share prices, the differences in monthly growth rates are of stochastic nature, which is confirmed by the stationarity test of the given time series as indicated in Table14.

Table 2 P
-Values in the Phases of D-F Unit Root Test Source: Authors.

Table 4
Causality between the Movement of the BELEX15 Stock Exchange Index and the Movement of the BIRS Stock Exchange Index

Table 5
Causality between the Movements of the BIRS Stock Exchange Index and the Movements of the BELEX15 Stock Exchange Index PANOECONOMICUS, 2015, Vol.62, Issue 4, pp.469-491

Table 6
BELEX15 Time Series

Table 7
ADF Unit Root Test (p-values)

Table 8
Cointegration Test

Table 9
Regression Model of the AIK Share Price Movements

Table 10
Regression Model of the BELEX15 Index Movements

Table 12
ECM for BELEX15 Index

Table 13
Statistical Significance of the Time Series Breaks for the Difference in Monthly Growth Rates of AIK Share Prices and BELEX15 Index PANOECONOMICUS, 2015, Vol.62, Issue 4, pp.469-491