3

Contagion Effects of Macroeconomic News Announcements on CEE Banks♦

Terhi Jokipii♣ Institute for International Integration Studies Trinity College Dublin Brian Lucey Business School and Institute for International Integration Studies Trinity College Dublin June 2005 Abstract

This paper examines banking sector return co-movements between the three largest Central and Eastern European countries that have recently joined the European Union (EU), the Czech Republic, Hungary and Poland. Defining contagion as a significant increase in cross-country correlations after a shock to one market, we assess the effect of the 1994-1996 Czech banking crisis and find that correlations between these markets increase significantly during the turbulent period. Adjusting for market volatility during this time, we find further evidence of contagion between the Czech Republic and Hungary, however we discover that movements between the Czech Republic and Poland are rather driven by strong real linkages through economic fundamentals. In order to gain additional understanding of these movements, we construct a set of dummy variables based on daily macroeconomic news announcements and test for the impact of domestic, regional and international data releases on the banking sectors of these markets. We find that regional news has a larger impact on banking index movements than international releases. In particular, we see that the index movements resulting from terms of trade releases are the most significant. At the international level, current account, terms of trade and unemployment figures have the greates effect. Almost all news released by the EU is highly significant in all time periods. No significant results are obtained for news released in the US. Key words: Contagion, Macroeconomic news, Banking sector, Stock returns JEL classification numbers: F30, F40, G15

♦ We would like to thank the Government of Ireland through the Program for Research in Third Level Institutions (PRTLI) for financial support for this project. Jokipii also acknowledges support from Helsingin Sanomat 100-year fund. We would futher like to thank Lieven Baele and the audience of the Global Finance Conference for their very useful comments and suggestions. ♣ Corresponding Author: Jokipiit@tcd.ie; Institute for International Integration Studies, Trinity College Dublin, College Green, Dublin-2; +353- 1-608-3209

4

I Introduction

Banking crisis is hardly a new phenomenon. In fact, the development of the international banking sector has consistently been marred by failures, generating a diverse array of mechanisms to reduce their strength and impact. In more recent times, the incidence of banking and financial sector crises has intensified and their effect on the domestic and international economy has become even more profound. Understanding both the nature and the causes of dramatic co-movements, as well as the cross-country and cross-market transmission of shocks with a view to gaining insight into market linkages, has consequently become of major empirical and analytical interest in international finance. Much of the academic literature in this field has focussed on analysing the extent to which financial spill-overs to both mature as well as to emerging markets exist in order to identify channels of transmission of both positive and negative shocks to foreign countries. With the recent EU enlargement, much of this attention has shifted towards the Central and Eastern European Countries (CEEC). However, to date, only very few empirical papers have focussed on the existence of contagion both to and between these countries. Literature has generally examined the effect of either the Russian, Asian or Czech crisis of the 1990’s on market co-movements. (See for example Fries, Raiser & Stern, 1999; Murzuch and Weller, 2000; Darvas & Szapáry, 2000; Gelos & Sahay, 2001, Wälti; 2003). This paper serves to understand the relationships that exist between the banking sectors of the CEECs by analysing the 1994-1996 bank crisis of the Czech Republic. We aim to determine whether there is evidence in favour of banking sector contagion during this time, or whether co-movements could rather be attributed to strong real linkages. Applying the traditional correlation analysis as a test of contagion, we examine the nature of the relationships between banking sectors of these economies both during crisis as well as post crisis. Defining contagion as a significant increase in cross-country correlations after a shock to one country, we are able to infer the existence of contagion. We further follow Forbes and Rigobon (2002) and adjust for market volatility during the turbulent period. Our findings indicate that evidence of contagion between the Czech Republic and Hungary can be inferred, while the high correlations between the Czech Republic and Poland are rather driven by strong real linkages. Finally, we create a set of dummy variables based on daily macroeconomic news announcements to test for the impact of domestic, regional and international data releases on the banking sectors of these markets. Only very few authors have attempted to use news as the identifying condition for the propagation of shocks. Eichengreen, Rose and Wyplosz (1997) study the collapse of fixed exchange rates in the ERM at the end of 1993, with one country’s collapse taken as the exogenous event. They subsequently calculate the probability of a country’s crisis affecting the probability of other countries facing a similar crisis. Baig and Goldfajn (1999) study the impact of daily news in one country’s stock market (exogenous event) on other countries’ markets during the East Asian crisis. To our knowledge no study has yet investigated the impact on news releases on the movements of CEE banking sector indices.

5

We find that it is a combination of domestic and regional news that is driving these movements, rather than international data releases. In particular, we find that at the regional level, releases relating terms of trade as well as to the dummy considering the countries’ credit rating, its move towards EU membership as well as other factors signalling greater development, are of most significance. For the international dummy variables, we find that almost all news released in the EU at all time periods are significant, while CPI and terms of trade figures from the UK had the greatest impact. We found no significant results when considering news released from the US. We further follow Favero and Giavazzi (2002) who model increases in volatility during the crisis period in one market by extreme movements in the banking sector returns of another market. The evidence here is slightly more mixed. The remainder of the paper is organized as follows: Section II introduces the concept of contagion and briefly looks at the literature in this field. In Section III, we briefly describe the evolution of the CEE banking systems, focussing on the development of the 1994 bank crisis. In Section IV, we describe the data employed in our analysis. In Section V we present our empirical methodology and results for testing for contagion. In Section VI, we try to determine the nature of propagation. Section VII briefly concludes. II Contagion

Contagion has been defined in many different ways in the literature, including the transfer of any shock across countries (Edwards 2000). Eichengreen and Rose (1999) and Kaminsky and Reinhart (1999) define contagion as the situation where the knowledge of crisis in one country increases the risk of crisis in another country. Edwards (2000) goes on to restrict the term economic contagion to those situations where the magnitude with which a shock is transmitted exceeds what was expected on the ex ante basis of ‘fundamentals’. In this paper we follow Forbes and Rigobon (2002) and consider contagion as the significant increase in cross-market correlations during a period of turmoil. Much of the empirical work on measuring the existence of contagion is based on comparing correlations between markets during a relatively stable period with a crisis or turbulent period (King & Wadhwani, 1990; Boyer, Gibson & Loretan, 1999; Loretan & English, 2000; Forbes & Rigobon, 2002; Corsetti, Pericoli & Sbracia, 2002). According to this approach coupled with the definition adopted in this paper, if two markets are naturally moderately correlated during periods of stability, then a shock to one market will result in a significant increase in market co-movements. This increase constitutes contagion. On the other hand, if the correlation does not increase significantly after a shock to one market, then continued market co-movements can be inferred to be driven by strong real linkages between the two economies, something Forbes and Rigobon (2002) define as interdependence. Based on these assumptions, contagion in this sense implies that cross-country linkages are fundamentally different after a shock to one market while interdependence implies no real change to relationships.

6

The theoretical literature on banking contagion has generally focussed on the transfer of shocks through the interbank market. Allen and Gale (2000) show that the possibility of contagion depends strongly on the completeness of the structure of interregional claims. They find that complete claims structures are shown to be more robust than incomplete structures. Freixas, Parigi and Rochet, 2000 investigate the ability of the banking sector to withstand the insolvency of one bank and whether the closure of one bank generates a rippling effect throughout the system. They find that contagion arises from unforeseen liquidity shocks i.e. Banks withdrawing interbank deposits from another bank. Gropp and Vesala, 2004 examine the number of banks that have in a given country to experience a large shock in the same period via the use of co-exceedances. They find significant evidence of cross-border contagion in the EU, attributing their findings to cross-border interbank exposures. III Central and Eastern European Banking Systems

Having endured over a decade of substantial reform and stabilization, banking and financial sector development in the Czech Republic, Hungary and Poland finds itself in its final stages. The privatization of large banks has ultimately been completed, and foreign strategic owners, most of them EU-based banks, dominate the banking sector. After an enormous clean up of portfolios, the standardization of banking sector regulations according to EU rules along with the new ownership structures, the Czech Republic, Hungary and Poland are finally facing the same issues challenging most industrialized nations. The Czech Republic, Poland and Hungary all encountered similar structural problems during the development of their banking systems. The Czech Republic however was additionaly faced with problems relating to the dissolution of the federal country. First, the CEECs had inherited underdeveloped, undercapitalized and badly managed banks from the past, having a strong impact on developments during the 1990s. At the outset of the transition, several key reforms were implemented. A two-tier banking system with separate functions for the central bank and commercial banks was introduced in place of the traditional mono-bank system. Furthermore, privately owned banks were admitted, and foreign banks and joint ventures were granted access. The licensing policy for most kinds of banking business was liberalised and both the legal and the supervisory system were adjusted. Together the largely liberal licensing policy and the unreliable legal framework and supervisory system, resulted in the establishment of a large number of newly founded banks engaged in unsound practices1. The state owned commercial banks which were borne from the mono-bank system, subsequently suffered from an inherited burden of bad loans. The banking system generally lacked capital and banking skills, and political intervention in the activities of state owned banks was persistent. These deficiencies, coupled with the uncertain economic 1 The cases of Komercní Banka and credit unions in the Czech Republic serve as examples in this respect.

7

environment culminated in the quick accumulation of bad loans, and the subsequent failure of around a dozen banks. Some statistics relating to the evolution of the banking sectors of the Czech Republic, Hungary and Poland are presented in Table I, Table II and Table III. Studies that have focussed on assessing the effect of the Russian crisis on CEECs have found that there is very little evidence that CEECs were affected through financial linkages. (See Darvas & Szapáry, 2000; Krzak, 1998; Lucey & Voronkova, 2005 for more detailed discussion in this field.) Accordingly, for our analysis we concentrate on the Czech banking crisis and consequently define our crisis phase as the period between 1994 and 1996. IV Data

For our empirical analysis, we make use of nine and a half years of daily data (July 1994 to end 2004). We adopt the DataStream banking and financial sector indices2 for the Czech Republic, Hungary and Poland, as well as for the United Kingdom (UK), United States (US) and EU. Adopting DataStream indices provides a comparative index for each country, which captures more than 75 per cent of the total market. The DataStream indices are value-weighted indices based on key publicly quoted banks in the each country. The weighting is based on the market capitalization of each bank. The larger the market cap, the greater the weighting, giving an accurate reflection of the banking sector. These indices were obtained in US dollar terms; their per cent changes and natural logs were subsequently calculated. The US series employed was lagged by one day in order to account for time differences. Creation of Dummy Variables

As no high frequency variables (e.g. daily data) that can approximate the fundamentals in each country exist for our estimations in Section V, we employ an approach whereby we create a set of dummy variables constructed from macro-economic news announcements. Such dummy series were created via the collection of daily news releases either from the central bank or national statistical office for the UK, EU3 and US, or where official figures were unavailable we employed the LexisNexis news databank.4 The searches were conducted focusing both on central bank and national statistical office releases as well as on the individual variables we wished to include. The variables of interest were: unemployment, short-term interest rates, consumer price index, terms of trade (or an indicator of market openness when not available), real gross domestic product (GDP), current account deficit. These indicators are largely

2 Analysis was conducted on both the bank and financial sector indices for all countries, however due to the largely similar results obtained only the banking sector results are presented. This similarity is hardly surprising considering that between 85-88% of total assets of the financial system are held by banks for all three countries under analysis. 3 The EU variable was created by combining data for Germany before 1999 with EU data after 1999. 4 Our sources were limited due to lack of alternatives on LexisNexis. We made use of MTI Econews for Hungary, CTK for the Czech Republic and the Polish News Bulletin for announcements in Poland.

8

considered to be of most importance in the assessment of macro-economic influence on the banking system stability in the literature (Huchinson, 1999; Demirgűç-Kunt & Detragiache, 1998; Timmermans, 2001). For Hungary, Poland and the Czech Republic, we also included various other announcements; such as a change in exchange rate regime, a change in the country’s credit rating by a major rating agency, namely Standard & Poor’s, Moody’s or Fitch, or the decision by the central bank to intervene in the foreign exchange market. Following Baig and Goldfjn (1999), we distinguished between ‘good’ and ‘bad’ news by using simple guidelines - any credible attempt to restructure or improve the economic situation was deemed as ‘good’, whereas news that indicated a further decline of the real or financial sector was deemed ‘bad’. The good (bad) news dummy series was assigned a ‘1’ on the release of favourable (unfavourable) macro-economic news. A fall in inflation, better GDP growth, and an improvement in the terms of trade were all assigned a ‘good’ news dummy, as were a fall in the consumer price index, lower unemployment figures, and a reduction in the current account deficit. Finally, we assumed that the central banks followed a price stability or inflation targeting strategy and assigned a ‘good’ news dummy to a decrease in the short-term interest rate, as set by the central bank. For the Czech Republic, Hungary and Poland, we further considered any news of the country’s move towards EU membership as ‘good’, together with an increase in the country’s credit rating by one of the major ratings agencies. Furthermore, we considered any move in exchange rate regimes towards a free-float as being positive and any rescue packages or funding given to the country by international organizations as favourable. While the privatization of state-owned banks can in principle have benefits as well as costs, the overall general considerations of the incentive effects of private ownership, as well as empirical evidence on this issue support the view that the benefits outweigh the costs (EBRD, 1997). Private ownership generally provides better incentives for more disciplined risk taking of managers along with the limitation of government intervention into the allocation of credit. Furthermore, it enhances the incentives for more effective monitoring and screening of banking institutions, ultimately improving the general stability and soundness of the banking system in which they operate. For this reason, we further considered any increase or speeding up of the privatization process as being positive. For ‘bad’ news the opposite was true. We assigned a ‘bad’ news dummy variable when the rate of inflation increased, GDP growth declined or the terms of trade index worsened. Furthermore, a rise in consumer prices, an increase in unemployment or a fall in the interest rate was considered bad for the economy. Although very general, we assumed that a rise in interest rates would signal non inflationary pressures and concerns of deflation resulting in notably below trend GDP growth and consequently a worsening economic situation. Again, for the Czech Republic, any delay in EU membership was considered ‘bad’, as was a fall in the country’s credit rating. We further considered a delay in the privatization process for banks operating in the Czech Republic, Hungary or Poland, as negative. Once all the good and bad news series were created for each variable separately, we aggregated the series to obtain a ‘good’ and a ‘bad’ news dummy series for each country. These aggregated series were employed for our regression analysis in Section VI.

9

V Testing for Contagion

Banking Sector Correlations

In order to obtain evidence on the extent of banking sector co-movements in these countries, we begin by estimating the unadjusted correlation coefficients of the daily changes in the banking sector indices of Hungary, Poland and the Czech Republic. The sample period begins on the 19th of July 19945 and spans till the end of December 2004. We estimate the correlations in the overall sample period, and subsequently apply the Likelihood ratio test for the significance of group-wise correlations following Baig and Goldfajn (1999), Valdés (1997), as well as Pindyck and Rotemberg (1990) testing the null hypothesis of no group-wise correlation.6 The unadjusted correlations are calculated according to the standard definition of the correlation coefficient:

yx

yx

σσ

σρ

,=

(1.)

Where xσ and yσ represent the variances of the daily changes in banking sector and total

market indices of two distinct countries. The results for the total sample banking sector cross-country correlations are presented in Table IV. All pairs demonstrate positive coefficients ranging between 0.30 for Hungary and the Czech Republic, and 0.25 for the Czech Republic and Poland. It is interesting to note that in the full sample correlation analysis all pairs of correlations appear to be unaffected by the significant turmoil that affected the Czech Republic. We further split our data into a crisis and a post crisis period to allow for a comparison between a tranquil and a turbulent period.7 Again we calculate the correlation coefficients. The results are presented in Table IV. At a first glance, we see that the crisis period correlations for all pairs are larger when compared to both the total sample and post crisis period uncovering evidence in favour of contagion. It is interesting to note that while only slightly, the correlation coefficient between the Czech Republic and Poland remains the lowest of the pairs at 0.27, while the coefficients for the Czech Republic and Hungary, and Hungary and Poland are very similar in magnitude, 0.32 and 0.31 respectively. Post crisis, the coefficient between the Hungary and Poland experiences the greatest fall (80 per cent) to 0.06, while the correlations between the Czech Republic and Hungary and the Czech Republic and Poland fall around 65 and 55 per cent respectively. 5 The first date on which data was available for all three markets.

6 The test statistic: RN log− is distributed as 2x with )1(21

−

pp degrees of freedom, where│R│is the

determinant of the correlation matrix, N, the number of observations and p the number of series being tested. 7 Crisis period stems from 19th July 1994 to end 1996. The post crisis period covers January 1997 to end 2004.

10

Rolling Correlations

The problem that exists with applying the full sample for correlation analysis is such that it smoothes out much of the shorter duration interactions between markets. For instance, while it is evident from the fall in correlation that events in the Czech Republic during the crisis period had a substantial impact on its neighbouring Hungary and Poland, it is impossible to disentangle these movements as they become diminished via the use of the entire sample. To further investigate co-movements during the crisis, we repeat the exercise for sub-samples consisting of three-month windows, rolling them till the end date. We calculate rolling correlations covering the crisis period only the results for which are presented in Table V. It is interesting to note that all pair-wise correlations experience a large degree of volatility throughout the period of crisis, with each resulting in negative coefficients at times. Unsurprisingly, the correlations between Hungary and Poland appear to be relatively stable and steadily correlated when compared to each countries individual cross-correlation with the crisis stricken country. The coefficients for the Czech Republic and Poland seem to be the most volatile, particularly between February and July 1995 when the correlation falls from 0.2 to -0.27. This pair-wise correlation shows signs of stabilization already around February to April of 1996. Looking more closely at the relationship between the Czech Republic and Hungary during the crisis period, we find a high degree of volatility, which doesn’t appear to settle completely before the end of 1996. The correlations fluctuate between positive and negative coefficients for much of the time even if the degrees of fluctuations are smaller than those observed at times between the Czech Republic and Poland. Likelihood Ratio Test

While the full sample and the rolling correlations help to determine the patterns of co-movements between markets, they are not useful for gaining an understanding as to whether there is a significant difference in these market correlations during turbulent and tranquil times. We follow Forbes and Rigobon (2002) whereby we try to determine whether market movements can be attributed to contagion or to interdependence. To address this issue, we apply a two-sample, heteroskedastic t-test, and examine whether a significant increase in correlations in the crisis period is evident. If the correlations increase significantly, then there are grounds for believing that these markets have moved away from relationships dictated by traditional movements of fundamentals. On the other hand, if the increases are not significant, then it is possible to assume that these markets are simply reacting to shocks that are common-cause or spill over generated. Applying the likelihood ratio test, we examine the null hypothesis that the post crisis period correlations are greater or equal to the correlation during crisis against the alternative hypothesis that cross-country correlation is greater during the crisis period:

1,

0,0 : jijiH ρρ ≥

11

where tji ,ρ is the correlation coefficient between country i and country j over period t. The

post crisis period is denoted ‘0’ and crisis period as ‘1’. The correlation coefficients are transformed using a Fisher transformation, so that they are approximately normally distributed with a mean tµ and variance 2

tσ :

−

+= t

ji

tji

t,

,

11

ln21

ρ

ρµ

(2.)

and

312

−

=

tt n

σ

(3.)

The test statistic is derived as follows:

21

1

21

0

20

10

+

−=

−−

nS

nS

XXU

(4.)

where tX and 2tS are the estimated sample mean and variance following the Fisher

transformation. The test statistic follows the t-distribution. The degrees of freedom are calculated as follows:

( )( ) ( )

1/

1/

//

1

21

21

0

20

20

20

200

20

−

+

−

+

nnS

nnS

nSnS

(5.)

The results for the heteroskedastic t-test are presented in Table VI. For all pairs, the crisis period correlations are greater than the corresponding post crisis period coefficients within a 1 per cent level of significance. Due to these fundamental differences in cross-country linkages, we can infer the existence of contagion for all pairs during the period under analysis. The Log likelihood ratio test statistics (LR) are presented at the top of each table respectively. The test aims to reject that all pair wise correlations are zero; we find evidence of this at the 1 per cent level of significance for all tests. Bias in the Correlation Coefficient

While the simple test of contagion based on correlation analysis above appears to show evidence of contagion between both the banking sector indices of the CEEC economies, recent literature has highlighted a bias in the correlation coefficient central to this analysis. (Forbes & Rigobon, 2002). This bias proves to be especially large during periods of market turmoil, which, in effect, is the focus of this test.

1,

0,1 : jijiH ρρ <

12

The use of the unadjusted (or conditional) coefficient assumes the bilateral analysis of markets, x and y, and a division of the data sample into a high variance (h) and a low variance (l) group, representing the crisis and post crisis periods respectively. We start by estimating the correlation coefficient of the crisis period according to the standard definition (1.). The results are summarised in Table VII. As discussed, it is possible that the increases in the correlation coefficients result from a bias due to market volatility during this time rather than purely from contagion. To assess whether this bias affects our tests of contagion we repeat the analysis, this time adjusting the correlations to account for higher volatility of returns during periods of turmoil. The adjustment takes the form:

( )22,

2,

2, 1

)(1 y

ix

ixiy

yy

ρσ

σσ

ργ

−

−+

=

(6.)

where yρ is the crisis period correlation between x and y, 2,ixσ and 2

,iyσ are the variance of

asset returns in the crisis and post crisis periods of the crisis country’s asset returns respectively.

31

31

11

ln21

11

ln21

−+

−

−

+−

−

+

=

xy

y

y

y

y

TT

FRρ

ρ

γ

γ

(7.)

Where xρ is the correlation between x and y during the post crisis period, yγ is the

corresponding correlation coefficient in the turbulent period as obtained above. These results are presented in Table VIII. In adjusting for an increase in market volatility, we find a significant difference in the results obtained. At the 5 per cent level of significance, we are again able to identify the existence of contagion between the Czech Republic and Hungary. Considering the Czech Republic and Poland however, while these markets appear to be highly correlated, this co-movement appears to be driven by strong real linkages through economic fundamentals rather than through the existence of contagion. Discussion

In order to gain some indication of the reasons for the distinct results observed in the correlation analysis, we take a closer look at some of the differences that existed between the banking sectors of Hungary in Poland between 1994 and 1996. Looking at previous research by Barth, Caprio and Levine (2001)8 relating to the effect of regulation and ownership on performance and stability, we try to briefly assess some possible 8 The data is obtainable from www.worldbank.org/research/interest/intrstweb.htm

13

reasons for result obtained above. Much of the data presented here is for 1999 but nevertheless gives an indication of the differences between the banking sectors of these economies. Capital regulation Barth, Capiro and Levine (2001) construct a capital regulatory index based on restrictions on the leverage potential for capital on the one hand, and on the sources of funds that are counted as regulatory capital on the other. The index is constructed so that higher values of the can be interpreted as greater stringency. The index ranges between 0 and 8. Chart 1 indicates that Poland has the most stringent capital regulation, while the Czech Republic has the lowest index value. Interestingly, the figure for Poland was exactly at the lower quartile of the EU group.

Chart 1: Capital Regulatory Index

0 1 2 3 4 5 6

czech republic

hungary

poland

source: Barth, Caprio & Levine, 2001 Supervision The index constructed to measure the formal supervisory power tries to describe the legal possibilities of supervisors to prevent and correct problems in the banking industry. Capturing formal power to take prompt correctative action, to restructure and reorganize a troubled bank or to declare insolvency, the index ranges from 0 to 16, with higher values indicating more supervisory power. The results shown below indicate that Hungary has almost maximum supervisory power, however both the Czech Republic and Poland don’t lag so far behind.

14

Chart 2: Official Supervisory Power Index

0 2 4 6 8 10 12 14 16

czech republic

hungary

poland

source: Barth, Caprio & Levine, 2001 We further include some indication of the average number of supervisors per bank as seen below.

Chart 3: Number of Supervisors per Bank

0 0.5 1 1.5 2 2.5 3

czech republic

hungary

poland

source: Barth, Caprio & Levine, 2001 It is interesting to note the distinct differences that exist to those observed in Chart 2 above. Foreign ownership The graph below gives an indication of the per centage of total assets owned by foreign banks. The figures here consider the fraction of the banking systems assets that are 50 per cent or more foreign owned.

15

Chart 4: % Total Assets Foreign Owned

0 10 20 30 40 50 60 70

czech republic

hungary

poland

source: Barth, Caprio & Levine, 2001 Hungary clearly dominates the countries we are studying with as much as 60 per cent of total assets under foreign ownership. Restrictiveness Finally, we present a measure of relative restrictiveness on a banks activities and ownership. The index ranges from 1 to 4, with the higher value demonstrating the higher degree of restrictiveness. For the purpose of the index, in terms of activity, a banking sector is considered unrestricted when a full range of activities can be conducted directly by the bank. If the activity cannot be directly conducted either by the bank or its subsidiaries, then it is considered restricted. Similarly for ownership, an unrestricted bank can own 100 per cent of the equity of any financial/non financial firm, while a restricted bank is inhibited from acquiring equity investment. We see that this measure is largely similar between the three CEECs, with Poland only slightly more restrictive in its business.

Chart 5: Overall Restrictiveness of Bank Activities and Ownership

0 0.5 1 1.5 2 2.5 3

czech republic

hungary

poland

source: Barth, Caprio & Levine, 2001

16

Furthermore, we find that between 1992 and 1995, Poland was faced with sustained development of effective banking supervision. New accounting principles were introduced in 1995 and prudential requirements were imposed. Furthermore, internal audit and controls were strengthened and the Bank Guarantee Fund commenced operation in 1995. A combination of these created vast improvement in the operating and regulatory environment of the banking sector of Poland between 1992 and 1995. Ownership structures additionally changed dramatically during this period as seen in Chart 6 below:

Chart 6: Change in the Ownership Structure of Polish Banking: 1993-1996

0 % 20 % 40 % 60 % 80 % 100 %

co-operative banks

foreign banks

private banks

state-owned banks

1996 1993source:National Bank of Poland It is possible that Poland's immunity to contagion from the Czech Republic can be attributed to its strong relations with banks in developed markets, or due to a combination of foreign ownership or supervisory environment. Further research into this could be fruitful in understanding what causes a banking sector to be more vulnerable to crisis than another. Vector Autoregression Analysis

As we have seen in the previous sections, the 1990s were generally marked by periods of chaos in the Eastern European countries when economic uncertainty together with a general lack of capital and skills in the banking sector ultimately resulted in a banking crisis. The Czech Republic in particular suffered from various banking failures/ defaults having a significant negative impact on its neighbouring Hungary and Poland. In order to try to understand the patterns of banking sector pressure, we make use of the Vector Auto-regression (VAR) methodology as it recognizes the endogeneity of all the variables in the system. Furthermore, it allows us to incorporate lagged values of the variables and to move away from contemporaneous correlations.

17

The VAR model takes the form of: ttt uByxLA =+)( (8.)

with p

ps LALALALA −−−−= ...1)( 21

,0)( =tuE ,),( 'Σ=tt uuE ,0),( '

=st uuE for ,st ≠ ,0),( =tt uxE ),,,,,( ttttttt CZbankPLbankHUbankEUbankUKbankUSbankx =

and )(Constyt =

In this standard VAR, x is a (1 x N) vector of variables, A is an (N x N) matrix of coefficients, u is an (N x 1) vector of white noise disturbance terms, and L represents the lag operator (for example, Lixt = xt-1). The x vector contains the daily change in the bank indices for the US (USbank), UK (UKbank), EU (EUbank), Hungary (HUbank), Poland (PLbank) and the Czech Republic (CZbank). This equation consequently allows us to analyze the full range of interaction between the banking sector indices for all of the six countries. We run the six variable VAR for each of the countries under analysis and obtain the impulse response functions for the shock originating from the given country. We include the banking sectors of the US, the UK and the EU in order to gain an indication of the degree to which regional shocks were responsible for banking sector movements. We chose a lag length of two according to both the Schwarz and Hannan-Quinn information criterion, and repeat the exercise for each of the countries. Testing for autocorrelation of residuals we find that all autocorrelations and partial autocorrelations at all lags are nearly zero, and all Q-statistics are insignificant with large p-values. This finding is consistent with no serial correlation in the residuals.9 Only shocks originating from the US appear to be significant for all three time frames, while the UK has a significant negative impact on Poland in the post crisis period. All shocks seem to be absorbed within four to five days.

9 These results are not presented here, but are available from the authors on request.

18

VI Determining the Nature of Propagation

The Impact of Macro-economic News Announcements

Domestic News

Following Baig and Goldfajn (1999), Ganapolsky and Schmukler (1998) and Kaminsky and Schmukler (1998) who estimate the impact of various news announcements on the movements of various markets, we expand our analysis further to attempt to discover whether the banking sectors of the Czech Republic, Hungary and Poland are receptive to macro-economic news released both domestically and internationally. In a first step, we run a simple regression including domestic good and bad news dummies for each of the three banking indices, with the objective of discovering the impact that good and bad news releases have on the indices during the total sample, crisis and post crisis periods. The regression takes the following form, and is run with robust coefficients controlling for heteroskedasticity:

Where ,1=i 3; Tt ,....1= and { } 2,10 ti N ϑε ≈

As the variables all demonstrate autocorrelation of order 1 the regressions were consequently run including two lags of the dependent variable. Including the lagged variables in the regressions however, proved to over-ride the effects coming from the news dummies. As a second step, we ran the regressions with the dependent variable set equal to the residual of equation (9.). The regression this time took the form:

∑ ++= eDummiest βγε (10.)

Where ,1=i 3; Tt ,....1= and { } 2,10 ti N ϑε ≈

The results were largely similar to those obtained under equation (9.) and hence only results excluding lagged variables are reported10. We further ran the regressions including upto 5 lags of the independent variables, however the results of our estimations remaind unchanged. A summary of the results for the domestic dummy regressions are presented in Table X, while

10Results obtained with the inclusion of lagged variables are not presented here, but are available from the authors on request.

∑ +++=− ttiti DummiesRR εββα 1, ( 9.)

19

the full results can be seen in Table XI: Panel-a, b, and c for total period, crisis, and post crisis period sub-samples respectively. In analyzing the total sample period (Table XI: Panel-a), we see that unfavourable domestic announcements had a significant downward effect on the banking sector of the Czech Republic. In particular, we see the banking sector index falling to unfavourable news releases relating to GDP and short-term interest rates. The Czech index further falls on both good and bad news released relating to 'other news'. The index of Hungary similarly fell on the release of unfavourable domestic GDP news, and appears to react negatively to the aggregated bad news dummy variable. We further see the Hungarian index rising on the release of favourable domestic GDP news as well as on both good and bad news relating to terms of trade. Finally in Poland, the banking sector index falls on good news relating to terms of trade, unemployment as well as to 'other news'. It further appears to rise significantly to favourable releases on short-term interest rates, current account as well as on bad news regarding GDP figures. The associated residuals from these regressions act as a further measure of contagion, this time controlling for fundamentals. Comparing the correlations of these residuals to those presented in Table IV, we see a substantial increase. The LR test demonstrates statistically significant group-wise correlations allowing us to conclude that contagion persists well above and beyond the identified fundamentals over the entire sample period. Turning to the crisis period (Table XI: Panel-b), we see that the Czech banking index falls significantly on unfavourable news released relating to the domestic current account, GDP as well as unemployment figures. The index further falls on the release of good 'other news'. During the crisis period, the Hungarian bank index experiences a significant reduction after negative current account, GDP as well as unemployment figures are announced. The index experiences further significant falls on negative 'other news' released, as well as on positive current account data. On the other hand, the Hungarian index rises significantly with positive GDP figures, as well as with both good and bad terms of trade data. The Polish banking sector index appears to fall significantly on the release of both domestic positive and negative terms of trade news, as well as with good and bad 'other news'. The index does however experience a significant rise on the release of positive short-term interest rate data. Again controlling for fundamentals, we see a substantial increase in the coefficients between all pairs of countries when compared to Table IV. The LR test statistic remains significant, revealing statistically significant group-wise correlations between the banking sector indices of these countries, thus providing further evidence of contagion during this period. Finally, for the post crisis period (Table XI: Panel-c), we find that the Czech banking sector index falls significantly on the release of negative GDP results as well as on the release of unfavourable short-term interest rates. Furthermore it reacts negatively to favourable 'other news'. Hungarys’ index on the other hand falls significantly on the release of poor GDP figures as well as to bad short-term interest rate news and 'other news'. The sectoral index further rises significantly on the release of good short-terms interest rates, terms of trade and unemployment figures. Furthermore, we see that the Hungarian banking index rises significantly with the aggregated good news dummy. Finally, in Poland, we see the banking sector index rise with

20

positive CPI figures as well as to favourable news regarding short-term interest rates. The index however falls on unfavourable domestic news relating to interest rates, 'other news' as well as with the aggregate bad news variable. The index further falls on favourable terms of trade news as well as with 'other news' released. We subsequently estimate the residual correlation matrix, which shows a substantial increase in all pair-wise coefficients when compared to those in Table IV. Statistically significant group-wise correlations are revealed via the LR test, implying the persistence of contagion effects between the analyzed markets. We can therefore conclude that during crisis and post crisis periods as well as over the total period, contagion persists between these markets above and beyond the identified fundamentals. International News

To extend the analysis further, we ran the regressions again, this time including the complete set of news dummies on the right hand side. These regressions serve to aid the understanding of the impact of cross-border news on respective banking sectors. The results are presented in Table XII-Panel a, b and c. In a first step regression, none of our results for the US were significant, we have therefore not presented them here.11 Looking first at regional effects, we see that over the entire sample period, good news relating to the Czech current account, has a significant positive effect on all three markets. Both positive and negative Czech terms of trade news have a further positive impact on all three markets. Both Hungary and Poland react significantly to 'other news' released in Hungary, with the indices rising on the release of good news, and falling on the release of bad news. This movement is hardly surprising considering the high degree of co-movement as outlined in Table IV. All three markets appear to be affected by Polish current account news, with both the Hungarian and the Polish banking indices rising significantly with good news, while the Czech index falls with bad news. All markets fall significantly with the release of favourable 'other news' from Poland, with the Czech index falling further with unfavourable 'other news'. Internationally, both good and bad news relating to the EU current account constitutes a significant fall in the indices of both Hungary and Poland. All three markets fall on bad news relating to the EU terms of trade, while Hungary and Poland appear to further fall significantly on the release of good news. The Hungarian banking sector index is further affected by unemployment figures, which constitutes a significant fall in the sectoral index. The Polsih index similarly falls on the release of favourable unemployment figures from the EU. The aggregated good and bad news variables both constitute a rise in the Hungarian index, with the bad news dummy slightly more significant. A similar result is observed for Poland. It is interesting to note that throughout the total sample period, the Czech Republic appears to be the least affected by news released in the EU. Turning to the UK, favourable CPI figures constitutes a significant fall in the banking sector indices of all three markets. Furthermore, the terms of trade variable has an impact on all

11 These are available from the authors on request.

21

three markets. The indices of both Hungary and Poland rise significantly on the release of bad news, while the Czech Republics’ index falls on the release of good news. Assessing the crisis period, we observe that all the Czech banking index reacts negatively to almost all news released. Bad ‘other news’ released in the Czech Republic constitutes a fall in the sector indices of all three countries. The index of the Czech Republic further falls significantly on the release of good domestic ‘other news’. Favourable GDP figures released in Hungary result in a rise in the domestic banking sector index, while negative GDP figures cause both the Czech and the Polish indices to fall significantly. Good news relating to Hungarian short-term interest rates cause both the domestic and the Polish indices to rise significantly, while the Czech banking index falls on the release of bad news in this respect. Hungarian terms of trade figures have a significant effect on all three markets. Good news result in a rise in the Hungarian and Polish indices, while bad news results in a significant fall in all three markets. Bad ‘other news’ from Hungary results in a fall in all three banking sector indices, and a simultaneous rise in the sectoral indices of Hungary and Poland. Current account data released from Poland has an impact on all three banking sectors. The indices of Poland and Hungary both rise on the release of good news, while the Czech index falls on the release of bad news. Negative news relating to Polish terms of trade figures constitutes a fall in all three banking sector indices. ‘Other’ Polish news has a further impact on all three sectoral indices, with good news resulting in a rise in both the Polish and the Hungarian indices, while the Czech index falls significantly. It is interesting to note that the indices of Hungary and Poland react almost identically throughout the crisis period to regional news releases, while the Czech index appears to fall on both positive and negative data announcements. Looking at international announcements, we find that current account data released in the EU has an impact on both Hungary and Poland. Good news results in a rise in both indices, while bad news constitutes a fall in the Hungarian index. EU short-term interest rates have an effect on all three markets. The Czech banking index falls on the release of good news in this respect, while the indices of both Hungary and Poland fall on the release of bad news. Good EU terms of trade figures further impact all three markets. The Czech index further falls on the release of this data, while the Hungarian and Polish banking sector index rise significantly. The Hungarian and Polish banking sector indices further fall with the announcement of negative terms of trade data. Unemployment figures have a similar effect. The Hungarian and and Polish banking sector indices rise significantly on the release of good news, and fall significantly on the release of bad news. The reactions to good news releases are slightly more significant in both cases. Favourable CPI figures released in the UK result in a significant fall in both the Czech and the Polish indices, while the Hungarian index rises. Unfavourable short term interest rate data results in a significant decrease of the Czech and Hungarian indices, while the Polish banking sector index appears to rise on the release of favourable interest rate news. Negative terms of trade figures relating to the UK result in a significant decrease in all three banking sector indices. During the post crisis period, we find that both good and bad 'other news' released in the Czech Republic has a significant negative impact on all three banking sector indices. All three

22

indices similarly fall significantly on the release of unfavourable Hungarian GDP figures, and all rise with the release of good news relating to Hungarian short-term interest rates. Terms of trade figures in Hungary result in a significant rise in the indices of all three countries, as does the release of good 'other news'. Bad 'other news' from Hungary however, negatively impacts all three markets. Polish CPI figures cause a significant positive move in all three banking sector indices. 'Other news' released in Poland has a further significant downward impact on all three markets. Looking at the effect of international releases, we find that good news relating to the EU current account results in a significant fall in all three markets, as does the release of favourable GDP figures. Positive EU terms of trade figures have a similar impact, as does good news relating to unemployment data. Good news relating to CPI from the UK results in a significant decrease in the banking sectors of all three markets. A similar effect is evident for GDP, terms of trade, as well as unemployment figures. Both good and bad news relating to short-term interest rates result in a significant fall in all three indices. Finally looking at the aggregate good news variable for the UK, we find that all three markets are positively and significantly affected. In order to test for stability in the parameters we have adopted for our analysis, we implement the CUSM test of stability, which is based on the cumulative sum of the recursive residuals; we find that all parameters demonstrate stability.12

Modelling the Volatility of Shocks

A further test of contagion is based on modelling increases in volatility during turbulent periods in one country’s index by extreme movements in the other country’s index. Following Favero and Giavazzi (2002), the process begins by considering a bivariate version (N=2) of the turbulent period. This reduced form model is simply a VAR specification, which takes the familiar form as presented in Section V. According to both the Schwarz and Hannan-Quinn information criterion the optimal lag length for our model is two. In order to identify country-specific shocks, we calculate the residuals tiu , and tju , of the

estimated VAR which are heteroskedastic and contain information on episodes of market turbulence. A dummy variable is subsequently assigned to the residuals, which lie above a certain threshold, in this case equal to three times the residual standard deviation of the VAR, capturing regional and country specific events. The choice of three standard deviations as a threshold is motivated by previous research. Gindraux (2000) provides a detailed sensitivity analysis for threshold values and finds that typically dummies selected with a threshold of three survive the testing-down procedure to over-identify the structural system of equations in the second step of the methodology. 12 These results are not presented here for brevity, but are available from the authors on request.

23

itiuotherwisetid σ3:1

:0,, >

⟨=

jtjuotherwisetjd

σ3:1:0,

, >

⟨=

(11.)

Residuals larger than the assigned threshold are considered as country-specific or local shocks. Dummy variables are created at the local (country-specific), regional (i.e. CZ,HU,PL or US,UK,EU) and global level (across all markets), and represent points in time when a combination of markets simultaneously experience extreme movements allowing for the eradication of hetroskedasticity and non-normality. The results obtained from the F-G test are presented in Table XIV. After running the reduced form VAR as described above, we identified 260 local shocks broken down as follows:

CZ HU PL EU UK US 48 26 61 43 39 43

It is interesting that Poland suffers the largest number of local shocks, while figures for the Czech Republic are almost double those seen in Hungary. On a more aggregated level, we were unable to identify any global shocks, or periods where turbulence existed simultaneously across the six countries. However, breaking the data into regional sections, we identified four events of extreme movements:

Eastern Europe (CZ, HU, PL) International (EU, UK, US) 2 2

Both shocks that occurred throughout Eastern Europe were negative in all three countries while one set of shocks at the international level proved positive and the other negative. A breakdown of the dates and types of shocks along with news announcements made on these days are presented in Table XIII. Inputting these shocks as dummies into the model, we test for contagion, the results of which are presented in Table XIV. Again, we find evidence of contagion between the banking sectors albeit to a differing extent and from varying origins. We find that local shocks originating in Poland, the EU and the US have a significant negative impact on the banking index of Hungary, while the international shock positively affects the Czech Republic.

24

Conclusion

Considerable co-movement appears to be present in the banking sectors of Hungary, Poland and the Czech Republic, indicating that the banking sectors of the largest markets in Central and Eastern Europe have tended to move together, and have had a substantial influence on one another. We consider the 1994-1996 Czech bank crisis and compare unadjusted (conditional) correlations between the crisis and post crisis periods. We are able to find evidence in favour of contagion during this time, as inferred by the significant in cross-country correlations. Adjusting for the increase in market volatility during the crisis period however, we find that contagion is only evident between the Czech Republic and Hungary. Despite the high correlations between the Czech Republic and Poland we are unable to find evidence of contagion and infer that such co-movements are attributable to close linkages through economic fundamentals. Finally estimating a VAR equation to determine the influence that domestic, regional and international banking system shocks have on these markets, we find that the US banking system has the largest consistent effect on the CEE markets, with these shocks generally absorbed within four to five days. We further create a set of dummy variables to try to capture the influence that macro-economic data releases have on the movements of these markets. We show that after controlling for own country news and other fundamentals, the cross-country correlations remain large and significant, and that consequently contagion exists between these markets. We find that macro-economic shocks originating both domestically as well as in neighbouring countries had the largest significant impact on these markets, with the EU becoming increasingly more important with time. In particular, we find that at the regional level, releases relating terms of trade as well as to the dummy considering the countries’ credit rating, its move towards EU membership as well as other factors signalling greater development, are of most significance. For the international dummy variables, we find that almost all news released in the EU at all time periods are significant, while CPI and terms of trade figures from the UK had the greatest impact. We finally model increases in volatility during the reform period in one country’s banking sector by movement in another country’s index. We again find evidence of contagion. Throughout our analysis, it is unclear whether the increase in importance of the EU markets can be attributed to integration, or merely due to the coincidence that the economic union was created at this time and has since become a major influence for these countries. As data becomes available, further research into this may be useful.

25

References

Allen, F., and Gale, D. (2000). "Financial Contagion." Journal of Political Economy 108(1): 1-33. Andersen, T., and Bollerslev, T. (1998). "Deutsche Mark-Dollar Volatility: Intra-day Activity Patterns, Macroeconomic Announcements, and Longer Run Dependencies." Journal of Finance 53: 219-265. Andersen, T., et al. (2003). "Micro Effects of Macro Announcements: Real-Time Price Discovery in Foreign Exchange." American Economic Review 93(1): 38-62. Bayoumi, T., et al. (2003). “Fatal Attraction: Using Distance to Measure Contagion in Good Times as Well as Bad.” Strathclide Discussion Papers in Economics 03-05. Baig, T., and Goldfajn, L. (1999). "Financial Markets Contagion in the Asian Crisis." IMF Working Paper 46(2). Baig, T., and Goldfajn, L. (2000). "The Russian Default and the Contagion to Brazil." IMF Working Paper 00/160. Barth, J., Caprio, G., and Levine, R. (2001). "The Regulation and Supervision of Banks Around the World: A New Database." World Bank Policy Research Working Paper 2588 (1). Becker, K., Finnerty, J., and Gupta, M. (1990). "The Inter-temporal Relation Between the U.S and Japanese Equity Markets." Journal of Financial Research 45: 1297-1306. Bennet, P., and Kelleher, J. (1998). "The International Transmission of Stock Price Disruption in October 1987." Federal Reserve Bank of New York, Quarterly Review 13: 17-33. Boyer, B., Gibson, M., and Loretan, M. (1999). "Pitfalls in Tests for Changes in Correlation." Federal Reserve Board, International Finance Discussion Paper 597. Brewer, E., and Jackson, W. (2002). "Inter-Industry Contagion and the Competitive Effects of Financial Distress Announcements: Evidence from Commercial Banks and Life Insurance Companies." FRB of Chicago Working Paper No. 2002-23. Calvo, S., and Reinhart, C. (1996). "Rational Contagion and the Globalization of Security Markets." Journal of International Economics 51: 79-113. Capiro, G., and Klingebiel, D. (1999). "Rational Contagion and the Globalization of Security Markets." Journal of International Economics 51: 79-113. Corsetti, G., Pericoli, M., and Sbracia, M. (2002). "Some Contagion, Some Interdependence: More Pitfalls in Tests of Financial Contagion." mimeo, University of Rome III. Darvas, S., and Szapáry, G. (2000). "Financial Contagion in Five Small Open Economies: Does the Exchange Rate Regime Really Matter?" International Finance 3: 25-51. De Brandt, O., and Hartmann, P. (2000). "Episodes of Systemic and Borderline Financial Crises." Mimeo World Bank.

26

De Gregorio, J., and Valdes, O. (2001). "Crisis Transmission: Evidence from the Debt, Tequila and Asian Flu Crises." The World Bank Economic Review 15(2): 289-314. Demirgűç-Kunt, A., and, Detragiache, E. (1998). "The Determinants of Banking Crisis in Developing and Developed Countries." IMF Staff Papers 45(1). Edwards, S. (1998). "Interest Rate Volatility, Capital Controls and Contagion." NBER Working Paper 6756. Edwards, S. (2000). "Contagion." The Wold Economy 23(7). Eichengreen, B., Rose, A., and Wyplosz, C. (1997). "Contagious Currency Crises." NBER Working Paper 5681. Eichengreen, B., Rose, A. (1999). "To Defend or Not Defend? That is The Question." Managing Currency Crises in Emerging Markets, edited by Michael Dooley and Jeffrey Frankel. European Bank for Reconstruction and Development. (EBRD): Transition Report 1997. London. Favero, C., and Giavazzi, F. (2000). "Looking for Contagion: Evidence from the 1992 ERM Crisis." Center for Economic Policy Research, Discussion Paper Series No. 2591. Favero, C., and Giavazzi, F. (2002). "Is the International Propagation of Shocks Non-Linear? Evidence from the ERM." Journal of International Economics 57(1), 231-46. Forbes, K., and Rigobon, R. (2002). "No Contagion, Only Interdependence: Measuring Stock Market Co-Movements." Journal of Finance 57, 2223-2261. Forbes, K. (2001). "How are Shocks Propagated Internationally? Firm-Level Evidence from the Asian and Russian Crises." in Reuven Click, Ramon Moreno and Mark M. Spiegel, eds. Financial Crises in Emerging Markets. Cambridge University Press: 106-159. Freixas, X., Parigi, B., and Rochet, J. (2000). "Systemic Risk, Interbank Relations and Liquidity Provision by the Central Bank." Journal of Money, Credit and Banking 32(3). Ganapolsky, E., Schmulker, S. (1998). "The Impact of Policy Announcement and News on Capital Markets: Crisis Management in Argentina During the Tequila Effect." World Bank. Gelos, G., and Sahay, R. (2000). "Financial Markets Spillovers in Transition Economies." IMF Working Paper 00/71. Gindraux, Y. (2000). "Contagion of Financial Crisis: Empirical Evidence from Russia and Brazil." Manuscript, Manuscript, Graduate Institute of International Studies. Glick, R., and Rose, A. (1998). "Contagion and Trade: Why are Currency Crises Regional?" NBER Working Paper 6806. Gropp, R., and Vesala, J. (2004). “Bank Contagion in Europe” mimeo ECB (December)

27

Hamao, Y., Masulis, R., and Ng, V. (1990). "Correlations in Price Changes and Volatility across International Stock Markets." Review of Financial Studies 3: 281-307. Huchinson, M. (1999). "European Banking Distress and EMU: Institutional and Macroeconomic Risks." Economic Policy Research Unit Working Paper 00-04. Kaminsky, G. S. (1998). "What Triggers Market Jitters? A Chronicle of the Asian Crisis." Policy Research Working Paper No. 2094: World Bank. Kaminsky, G., and Reinhart, C. (1999). "The Twin Crises: The Cause of Banking and Balance-of-Payments Problems." American Economic Association 89(3): 473-500. Kaminsky, G., and Reinhart, C. (2000). "On Crises, Contagion and Confusion." Journal of International Economics 51(1): 145-168. Kaminsky, G., and Schmukler, S. (1998). "What Triggers Market Jitters? A Chronicle of the Asian Crisis." Policy Research Working Paper No. 2049: World Bank. Kawalec, S. (1999). "Banking Sector Systemic Risk in Selected Central and Eastern European Countries- Review of Bulgaria, Czech Republic, Hungary, Poland, Romania and Slovakia.” Revised Draft, mimeo. Warsaw. King, M., and Wadhwani, S. (1990). "Transmission of Volatility Between Stock Markets." Review of Financial Studies 3(1): 5-33. Krzak, M. (1998). "Contagion Effects of The Russian Financial Crisis on Central and Eastern Europe: The Case of Poland." in Focus on Transition, Österreichiesche Nationalbank, 22-37. Lagunoff, R., and Schreft, S. (1999). "Financial Fragility with Rational and Irrational Exuberance." Journal of Money, Credit and Banking 31(3): 531-560. Lang, L., and Stulz, R. (1992). "Contagion and Competitive Intra-Industry Effects of Bankruptcy Announcements." Journal of Financial Economics 32(45-60). Linne, T. (1999). "Contagion Effects of Central and Eastern European Currency Crisis." Institut für Wirtschaftsforschung Halle 96. Lin, W., Engle, R., and Ito, T. (1994). "Do Bulls and Bears Move Across Borders? International Transmission of Stock Returns and Volatility." Review of Financial Studies 7: 506-538. Loretan, M., and English, W. (2000). "Evaluating Correlation Breakdowns During Periods of Market Volatility." In International Financial Markets and the Implication for Monetary and Financial Stability ed. Bank for International Settlements. Lucey, B., and Voronkova, S. (2005) "Linkages and Relationships Between Emerging, European and Developed Stock Markets Before and After the Russian Crisis of 1997-2000", Institute for International Integration Studies, Discussion Paper 54. Mishkin, F. (2001). "Prudential Supervision: Why Is It Important and What are the Issues?" in Prudential Supervision: What Works and What Doesn't eds, University of Chicago Press: 1-29.

28

Murzuch, B. and Weller, C. (2002). "International Financial Contagion: Why Are Eastern European Banks Not Failing When Everyone Else's Are?" Economics of Transition 8(3). Pindyck, R., and Rotemberg, J. (1990). "The Excess Co-movement of Commodity Prices." The Economic Journal. Rigobon, R. (2001). "Contagion: How To Measure It?" NBER Working Paper 8118. Staub, M. (1999). "Aspects of Systematic Risk in Banking: Inter Bank Loans, Optimal Bank Size and the Swiss Regional Bank Crisis." Basel. St.Clair, R. (2004). “Macroeconomic Determinants of Banking Financial Performance and Resilience in Singapore.” Monetary Authority of Singapore Working Paper 38. Tai, C. (2004a). "Contagion: Evidence from the International Banking Industry." Journal of Multinational Financial Management 14(4-5): 353-368. Tai, C. (2004b). "Can Bank be a Source of Contagion During the 1997 Asian Crisis?" Journal of Banking and Finance 28(2): 399-421. Timmermans, T. (2001). "Monitoring Macroeconomic Determinants of Banking system Stability." BIS Working Paper 1(Part 4). Valdés, R. (1997). "Emerging Market Contagion: Evidence and Theory." Documentos de Trabajo del Banco Central de Chile. Van Rijckeghem, C., and Weder, B. (2000). "Spillovers Through Banking Centers- A Panel Data Analysis." IMF Working Paper 00/88. Wälti, S. (2003). "Contagion and Interdependence Among Central and Eastern European Economies: The Impact of Common External Shocks." National Centre of Competence in Research, Financial Valuation and Risk Management, Working Paper 48.

Table I: Fiscal Costs of Bank Recapitalization

Source: IMF (1998), Kawalec (1999), National Central Banks

Table II: Evolution of Eastern European Banking Sectors

Source: National Central Banks

Table III: Banking Sector of the Czech Republic, Poland and Hungary

Source: National Central Banks

Czech Republic 1997

Hungary 1994

Poland 1996

Main part of the recapitalization program completed in fiscal costs up to the year indicated in % of GDP of that year

8.9% 7.2% 1.6%

Fiscal costs of the recapitalization program up to The year 2000 in % of GDP in 2000

11.8% 6.8% 1.4%

1994 1995 1996 1997 1998 1999 2000 Market value of banking index (US$) Czech Republic 1515 1138 1824 927 470 1433 Hungary 23 202 532 1058 1362 1627 Poland 998 1354 2849 3517 6152 8472 Foreign ownership (% of net assets) Czech Republic 23 24 30 39 48 55 Hungary 79 83 93 89 91 91 Poland 3 4 14 15 17 47 70

CZECH REPUBLIC HUNGARY POLAND 1991 1994 1997 2000 1991 1994 1997 2000 1993 1994 1997 2000 # Active banks 24 52 53 40 36 43 44 39 104 82 83 74 % Total domestic controlled

83 65 57 35 84 37 31 55 51 47 27

% Total foreign controlled

17 35 43 65 16 63 69 45 49 53 73

30

Table IV: Correlations of the CEECs Banking Index Returns

Total sample: LR Test 251.29 ***

CZ HU PL

CZ 1 0.30 0.25 HU 0.30 1 0.28 PL 0.25 0.28 1 Crisis (1994-end 1996): LR Test 68.94***

CZ HU PL CZ 1 0.32 0.27 HU 0.32 1 0.31 PL 0.27 0.31 1 Post crisis (1997- end 2004): LR Test 26.63***

CZ HU PL CZ 1 0.11 0.12 HU 0.11 1 0.06 PL 0.12 0.06 1

Note: *, ** and *** denote rejection at the 10%, 5% and 1% levels respectively. LR Test attempts to reject the null hypothesis that all pair wise correlations are 0. Each pair wise correlation is tested for the null that the correlation is 0.

Table V: Banking Sector Index Rolling Correlations

01/08/1994 to 31/10/1994: LR Test 2.87*** 01/11/1994 to 31/01/1995: LR Test 1.02***

CZ HU PL CZ HU PL CZ 1 0.02 0.31 CZ 1 0.13 0.13 HU 0.02 1 -0.01 HU 0.13 1 0.01 PL 0.31 -0.01 1 PL 0.13 0.01 1 01/02/1995 to 28/04/1995: LR Test 0.64*** 01/05/1995 to 31/07/1995: LR Test 3.05***

CZ HU PL CZ HU PL CZ 1 0.15 0.02 CZ 1 0.17 -0.27 HU 0.15 1 0.00 HU 0.17 1 -0.11 PL 0.02 0.00 1 PL -0.27 -0.11 1 01/08/1995 to 31/10/1995: LR Test 0.76*** 01/11/1995 to 31/01/1996: LR Test 2,78***

CZ HU PL CZ HU PL CZ 1 -0.05 -0.13 CZ 1 0.23 -0.02 HU -0.05 1 0.09 HU 0.23 1 0.19 PL -0.13 0.09 1 PL -0.02 0.19 1 01/02/1996 to 30/04/1996: LR Test 1.07*** 01/05/1996 to 31/07/1996: LR Test 1.74***

CZ HU PL CZ HU PL CZ 1 -0.11 0.11 CZ 1 -0.04 0.02 HU -0.11 1 0.10 HU -0.04 1 0.24 PL 0.11 0.10 1 PL 0.02 0.24 1 01/08/1996 to 31/10/1996: LR Test 5.24*** 01/11/1996 to 30/01/1997: LR Test 0.74***

CZ HU PL CZ HU PL CZ 1 0.37 0.15 CZ 1 0.02 0.12 HU 0.37 1 0.13 HU 0.02 1 0.11 PL 0.15 0.13 1 PL 0.12 0.11 1 Note: *, ** and *** denote rejection at the 10%, 5% and 1% levels respectively. LR Test attempts to reject the null hypothesis that all pair wise correlations are 0. Each pair wise correlation is tested for the null that the correlation is 0.

Table VI: Heteroskedastic t-Test: Banking Sector

Crisis (1994-end 1996): LR Test 68.94***

CZ HU PL CZ 1 0.32*** 0.27*** HU 0.32*** 1 0.31*** PL 0.27*** 0.31*** 1 Post crisis (1997- end 2004): LR Test 26.23***

CZ HU PL CZ 1 0.11*** 0.12*** HU 0.11*** 1 0.06*** PL 0.12*** 0.06*** 1

Note: *, ** and *** denote rejection at the 10%, 5% and 1% levels respectively. LR Test attempts to reject the null hypothesis that all pair wise correlations are 0. Each pair wise correlation is tested for the null that the correlation is 0. The Heteroskedastic t-test attempts to reject the null of equal correlations.

Table VII: Unadjusted Correlations

Relationship Crisis period

Post crisis period

Total sample

Test stat Contagion?

CZ-HU 0.32 0.11 0.30 12.65*** Yes CZ-PL 0.27 0.12 0.25 5.22*** Yes PL-HU 0.31 0.06 0.28 2.43*** Yes

Note: *, ** and *** denote rejection at the 10%, 5% and 1% levels respectively.

Table VIII: Adjusted Correlations

Contagion source

Country Actual crisis period

Adjusted crisis period

Test stat Contagion?

CZ HU 0.32 0.46 1.83** Yes CZ PL 0.27 0.44 0.09 NO

Note: *, ** and *** denote rejection at the 10%, 5% and 1% levels respectively. Table IX: Variance Decomposition

Days 1 2 3 4 5 6 7 8 9 10 The Czech Republic

Total period 0.0 0.7 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 Crisis period 0.1 1.0 1.8 1.8 1.8 1.9 1.9 1.9 1.9 1.9 Post crisis period 0.0 0.6 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5

Hungary Total period 0.0 0.8 3.3 3.4 3.4 3.4 3.4 3.4 3.4 3.4 Crisis period 0.0 1.9 6.6 6.7 6.8 6.8 6.8 6.8 6.8 6.8 Post crisis period 0.0 0.6 2.7 2.7 2.7 2.7 2.7 2.7 2.7 2.7

Poland Total period 0.0 1.1 6.2 6.4 6.4 6.4 6.4 6.4 6.4 6.4 Crisis period 0.0 0.8 10.3 10.7 10.7 10.7 10.7 10.7 10.7 10.7 Post crisis period 0.0 1.5 4.3 4.3 4.3 4.3 4.3 4.3 4.3 4.3

32

Table X: Summary of Own Country Dummies

Good News Bad News Panel-a: Total period

Czech Republic CPI Current account GDP -*** Short-term interest rate -*** Terms of trade Unemployment Other news -*** -*** Aggregated varible

Hungary CPI Current account GDP +*** -*** Short-term interest rate Terms of trade +*** +*** Unemployment -*** Other news -*** Aggregated variable

Poland CPI Current account +*** GDP +*** Short-term interest rate +*** Terms of trade -*** Unemployment -* Other news -*** Aggregated variable Panel-b: Crisis Period (1994 - end 1996)

Czech Republic CPI Current account -*** GDP -*** Short-term interest rate Terms of trade Unemployment -*** Other news -*** Aggregated variable

Hungary CPI Current account -* -** GDP +* -*** Short-term interest rate Terms of trade +*** +*** Unemployment -* Other news -*** Aggregated variable

Poland CPI Current account GDP Short term interest rate +** Terms of trade -** -*** Unemployment Other news -* -*** Aggregated variable Panel-c: Post crisis (1997- end 2004)

Czech Republic CPI Current account GDP -*** Short term interest rate -*** Terms of trade Unemployment Other news -*** Aggregated variable

Hungary CPI -** Current account GDP +* -*** Short term interest rate +*** Terms of trade +*** +*** Unemployment Other news -*** Aggregated variable +* -***

Poland CPI +** Current account GDP Short term interest rate +** -** Terms of trade -** Unemployment Other news -*** -*** Aggregated variable -*

Note: ***, ** and * denote significance at the 1, 5 and 10 % levels respectively. Only significant results are shown.

34

Table XI: Regression Results With Own Country Dummies

Panel-a: Total sample (1994- end 2004) CZ HU PL

Constant

1.72 (0.2177)

3.72 (0.8393)

20.00 (0.6053)

CPI Good news 0.16

(0.6015) -0.08

(0.1267) 1.50

(1.1244) Bad news -0.16

(0.9295) -0.70

(0.9972) -0.02

(0.0241) Current account Good news 0.31

(1.1755) 0.36

(0.5410) 2.14

(1.8168)* Bad news 0.03

(0.1520) -1.53

(1.5139) -0.16

(0.1557) GDP Good news 0.10

(0.4403) 1.75

(2.9163)*** -1.32

(1.1857) Bad news -0.86

(2.6830)*** -2.66

(4.5232)*** 3.92

(1.9917)** Short term interest rate Good news 0.12

(0.3282) -0.72

(0.8383) 3.17

(1.7800)* Bad news -0.53

(3.0616)*** 0.07

(0.0763) 1.99

(1.4810) Terms of trade Good news 0.01

(0.0747) 3.59

(5.6680)*** -4.35

(3.4836)*** Bad news 0.04

(0.2094) 5.98

(7.3591)*** 1.89

(1.6199) Unemployment Good news 0.12

(0.6552) -0.79

(3.0082)*** -2.58

(1.8178)* Bad news -0.16

(0.9437) 0.62

(1.0812) 0.38

(0.3098) Other news Good news -0.39

(2.9014)*** 0.42

(1.1140) -1.78

(3.4311)*** Bad news -0.41

(2.61568)*** -3.13

(4.5927)*** -0.77

(1.1546) Aggregated variable Good news 0.08

(0.6022) -0.44

(1.5551) 0.95

(1.7000) Bad news 0.02

(0.1130) 1.54

(2.5962)*** -0.36

(0.5603)

Adjusted R2 0.018 0.026 0.013 Number of Obs. 2729 2729 2729 Residual Correlation Matrix: LR Test 187.65*** CZ HU PL CZ 1 0.46 0.61 HU 0.46 1 0.82 PL 0.61 0.82 1 Panel b: Crisis period (1994- end 1996) CZ HU PL Constant

1.72 (0.1765)

4.07 (0.4543)

19.76 (0.5675)

CPI

Good News

-0.19 (0.1897)

-1.33 (0.3923)

1.39 (1.0232)

Bad News

-0.65 (0.2837)

2.98 (1.0287)

0.87 (0.8765)

Current account Good News

-3.29 (1.0003)

-0.98 (1.2897)*

2.48 (0.2882)

Bad News

-2.98 (2.9865)***

-1.64 (1.6874)**

-1.95 (0.7265)

GPD Good News

-1.25 (0.9875)

0.21 (1.4873)*

1.39 (1.0923)

Bad News

-4.67 (4.6125)***

-2.29 (2.9983)**

-0.75 (0.0843)

Short term interest rate Good News

-3.59 (0.2367)

1.49 (0.0983)

1.68 (2.0198)**

Bad News

-4.67 (4.6125)***

0.25 (1.0023)

-1.01 (0.3928)

Terms of trade Good News

-0.87 (0.0982)

2.37 (4.9821)***

-0.98 (1.9752)**

Bad News

-1.95 (1.0232)

1.86 (2.0875)**

-2.09 (3.0120)***

Unemployment Good News

-0.67 (1.0231)

1.45 (0.0010)

2.85 (1.0821)

Bad News

-3.62 (3.9821)***

-1.52 (2.3101)*

1.53 (0.1023)

Other news Good News

-2.97 (0.9976)

-0.96 (0.9201)

-1.94 (1.3241)*

Bad News

-4.04 (3.2531)***

-1.52 (2.3101)***

-2.64 (3.9723)***

Aggregate variable Good News

-0.98 (1.0231)

1.24 (1.0987)

3.29 (0.8764)

Bad News

-2.39 (1.0728)

-1.52 (0.6242)

-2.13 (1.2964)

Adjusted R2 0.021 0.024 0.015 Number of Obs. 640 640 640 Residual Correlation Matrix: LR Test 206.87***

CZ HU PL CZ 1 0.75 0.83 HU 0.75 1 0.82 PL 0.83 0.82 1 Panel c: Post crisis period (1997- end 2004)

CZ HU PL Constant

1.75 (0.1661)

4.67 (0.0065)

21.94 (0.5175)

CPI Good news 0.19

(0.6467) 0.66

(0.8131) 2.66

(2.4834)** Bad news -0.25

(1.2533) -1.53

(2.1201)** 0.51

(0.5143) Current account

36

Good news 0.27 (0.8642)

-0.30 (0.4373)

0.87 (0.6755)

Bad news -0.01 (0.2426)

-0.29 (0.2426)

-1.38 (1.3495)

GDP Good news 0.13

(0.4377) 1.17

(1.8882)* -0.45

(0.4100) Bad news -0.94

(2.7660)*** -3.84

(6.3096)*** 3.16

(1.3379) Short term interest rate Good news 0.19

(0.3157) 1.11

(3.4831)*** 2.73

(2.2027)** Bad news -0.57

(2.9152)*** -0.42

(0.4182) -2.82

(2.4796)** Terms of trade Good news 0.00

(0.0209) 2.59

(4.2913)*** -2.04

(1.9792)** Bad news 0.14

(0.6344) 5.04

(6.2228)*** 0.86

(0.7048) Unemployment Good news 0.32

(1.3267) -0.92

(3.0313)*** 0.14

(0.1224) Bad news -0.23

(1.1093) 0.82

(1.2914) 0.32

(0.2916) Other news Good news -0.53

(3.0707)*** 0.04

(0.0907) -2.43

(4.7758)*** Bad news -0.53

(2.8984)*** -3.22

(4.3435)*** -1.97

(3.1731)*** Aggregated variable Good news 0.09

(0.5408) 0.53

(1.6559)* 0.25

(0.4321) Bad news 0.06

(0.3306) -1.76

(2.8685)*** -1.12

(1.7765)* Adjusted R2 0.023 0.023 0.017 Number of Obs. 2087 2087 2087 Residual Correlation Matrix: LR Test 264.10*** CZ HU PL CZ 1 0.42 0.54 HU 0.42 1 0.60 PL 0.54 0.60 1

Note: *, ** and *** denote rejection at the 10%, 5% and 1% levels respectively. LR Test attempts to reject the null hypothesis that all pair-wise correlations are 0. Absolute value of t-statistics in parenthesis. National Bank Index set as dependant variable.

Table XII: Regression Results with International Dummies Panel-a: Total sample (1994- end 2004)

CZ HU PL Constant

1.49 (0.8502)

2.80 (0.3438)

19.81 (0.0809)

Czech Republic CPI Good news

0.10 (0.6157)

0.96 (1.8954)*

1.11 (0.6178)

Bad news

-0.10 (0.7610)

0.12 (0.3962)

0.61 (0.5905)

Current account Good news

0.29 (1.7006)*

0.93 (1.9054)*

3.18 (2.0447)**

Bad news

0.05 (0.3309)

0.46 (1.2659)

1.59 (1.3869)

GDP Good news

0.05 (0.3454)

0.34 (0.8078)

1.20 (0.9000)

Bad news

-0.62 (2.0191)**

0.70 (1.0545)

0.23 (0.1128)

Short term interest rate Good news -0.06

(0.2234) -0.87

(2.0140)** -1.17

(1.0456) Bad news -0.31

(2.0649)** 0.32

(0.9462) 0.18

(0.1489) Terms of trade Good news -0.05

(0.5563) -0.21

(0.8794) -0.74

(1.0116) Bad news 0.04

(0.2759) 0.18

(0.4659) 0.58

(0.4605) Unemployment Good news 0.15

(1.1272) 0.06

(0.1625) -0.23

(0.1880) Bad news -0.07

(0.5393) 0.18

(0.5757) 0.73

(0.7170) Other news Good news -0.14

(1.3648) -0.27

(1.0228) -1.08

(1.2398) Bad news -0.21

(1.5936) -0.07

(0.2434) -0.54

(0.5233) Aggregate news variable Good news 0.03

(0.3334) -0.13

(0.5639) 0.10

(0.1363) Bad news 0.01

(0.0564) -0.17

(0.5820) -0.41

(0.4140) Hungary

CPI Good news 0.18

(1.0707) -0.08

(0.1775) 0.96

(0.8322) Bad news -0.30

(2.1519)** 0.02

(0.0544) -0.17

(0.1296) Current account Good news -0.04

(0.2035) 0.36

(0.7959) 1.71

(1.2975) Bad news 0.03

(1.2608) -0.71

(2.2240)** -1.73

(3.9134)*** GDP Good news 0.03

(0.2395) 0.75

(2.1981)** 1.69

(1.5225) Bad news -0.87

(1.2608) -0.83

(2.2240)** -4.38

(3.9134)***

38

Short term interest rate Good news 0.28

(0.9629) -0.33

(0.4143) -0.28

(0.0967) Bad news -0.22

(1.2879) -0.27

(0.5547) 0.08

(0.0526) Terms of trade Good news 1.50

(4.3269)*** 3.69

(7.4865)*** 11.46

(8.8071)*** Bad news 1.61

(4.0469)*** 4.52

(20.3266)*** 10.24

(7.0104)*** Unemployment Good news 0.00

(0.0572) -0.16

(0.8106) -0.57

(0.8773) Bad news 0.09

(0.7746) 0.37

(1.1979) 0.81

(0.7994) Other news Good news 0.04

(0.4231) 0.65

(2.3919)** 1.96

(2.3419)** Bad news -0.26

(1.3479) -1.46

(2.8678)*** -5.32

(3.0760)*** Aggregated variable Good news 0.06

(0.7704) 0.08

(0.3978) 0.11

(0.1674) Bad news 0.20

(1.7115) 0.61

(1.6568)* 1.35

(1.2103) Poland

CPI Good news 0.18

(1.1142) 0.43

(0.9927) 1.33

(0.9934) Bad news 0.00

(0.0138) -0.11

(0.3664) -0.14

(0.1391) Current account Good news 0.18

(1.0416) 0.98

(2.1399)** 2.56

(2.1106)** Bad news -0.32

(2.5622)** 0.01

(0.0307) -0.29

(0.2670) GDP Good news -0.09

(0.6994) -0.44

(1.4532) -1.58

(1.5851) Bad news 0.14

(0.4118) 0.58

(0.9272) 3.81

(2.0631)** Short term interest rate Good news -0.51

(1.7465)* -0.32

(0.6727) -3.22

(1.7148)* Bad news 0.05

(0.3076) 0.11

(0.2912) 1.67

(1.2167) Terms of trade Good news -0.03

(0.1957) -0.99

(2.7005)*** -3.69

(2.8016)*** Bad news 0.02

(0.1297) 0.07

(0.1683) 2.26

(2.0475)** Unemployment Good news -0.08

(0.5447) -0.91

(2.2149)** -2.64

(1.7483)* Bad news -0.10

(0.6014) 0.35

(1.0428) 0.88

(0.7575) Other news Good news -0.28

(3.5189)*** -0.43

(2.7913)*** -1.73

(3.2950)*** Bad news -0.34

(3.3314)*** -0.14

(0.6880) -0.82

(1.2207) Aggregate variable Good news -0.02

(0.3299) 0.19

(1.0870) 0.75

(1.3615)

Bad news -0.14 (1.6777)*

0.10 (0.5368)

-0.36 (0.5621)

EU CPI Good news 0.03

(0.1137) 0.16

(0.3281) 0.37

(0.2222) Bad news -0.11

(0.4445) -1.39

(2.9412)*** -3.01

(1.4718) Current account Good news -0.09

(0.4846) -1.34

(3.6000)*** -3.72

(2.9231)*** Bad news -0.22

(0.8366) -2.12

(4.6608)*** -4.85

(2.2124)** GDP Good news -0.03

(0.1675) -0.64

(1.8337)* -1.53

(1.2589) Bad news 0.33

(1.0277) -0.83

(1.0313) -0.62

(0.2123) Short term interest rate Good news -0.49

(1.8800)* 0.12

(0.2878) -1.52

(1.4912) Bad news -0.02

(0.0725) -1.81

(2.4455)* -5.10

(1.8345) Terms of trade Good news -0.10

(0.5073) -0.67

(1.9301)* -2.41

(1.9561)** Bad news -0.43

(1.7495)* -1.40

(3.3530)*** -3.81

(2.0730)** Unemployment Good news -0.33

(1.5304) -1.46

(3.6216)*** -4.76

(3.4288)*** Bad news 0.14

(0.5133) -1.86

(3.6723)*** -3.21

(1.3888) Aggregate variable Good news 0.17

(0.0429) 1.83

(1.7654)* 4.16

(1.4182) Bad news 0.62

(0.6434) 0.36

(3.8771)*** 2.08

(1.9660)* UK

CPI Good news -0.53

(2.2653)** -0.94

(2.0801)** -3.41

(2.0234)** Bad news -0.22

(0.6048) -0.19

(0.3102) 0.77

(0.3896) Current account Good news -0.31

(1.6020) -0.76

(1.9967)** -2.00

(1.4922) Bad news -0.24

(0.6318) -0.95

(1.2875) -1.84

(0.7787) GDP Good news -0.57

(2.5821)*** 0.54

(1.1839) -0.12

(0.0702) Bad news -0.45

(1.3486) 0.32

(0.4698) 1.15

(0.5123) Short term interest rate Good news -0.52

(0.0725) -1.14

(2.4455)* -4.40

(1.8345) Bad news -0.65

(1.7486)* -0.41

(0.6394) -0.68

(0.3223) Terms of trade Good news -0.66

(2.9189)*** 0.71

(1.5316) 0.38

(0.2189) Bad news 0.00

(0.0084) 1.12

(1.8129)* 4.60

(2.3995)**

40

Unemployment Good news -0.70

(2.9644)*** -0.60

(1.1949) -2.64

(1.4247) Bad news -0.15

(0.3840) 0.21

(0.2881) 1.32

(0.5795) Aggregate variable Good news 0.62

(2.7220)*** 0.36

(0.7620) 2.08

(1.1729) Bad news 0.25

(0.6895) 0.39

(0.6237) -0.32

(0.1624)

Adjusted R2 0.070 0.079 0.066 Number of Obs. 2729 2729 2729 Panel-b: Crisis period (1994-1996)

Constant

1.45 (0.5674)

3.05 (0.3567)

21.83 (0.8263)

Czech Republic CPI Good news -0.15

(1.0282) 0.83

(2.0987)** 1.56

(0.9323) Bad news -1.62

(1.0403) -0.97

(0.9876) -2.83

(1.2932) Current account Good news -1.43

(0.0098) 0.93

(1.8372)* 1.29

(1.1092) Bad news -0.67

(2.3028)** -1.28

(0.8976) -2.18

(0.0987) GDP Good news -0.07

(1.0231) 2.03

(0.4938) 1.34

(1.8765)* Bad news -1.27

(1.9522)* -0.29

(1.1029) -1.29

(1.0029) Short term interest rate Good news -2.87

(0.9282) -1.87

(1.9872)* -1.12

(0.0928) Bad news -1.29

(1.2762) 0.82

(1.0243) 2.39

(1.0922) Terms of trade Good news 1.02

(0.9763) -2.62

(1.1293) -0.87

(1.5763)* Bad news -2.39

(1.2302) 1.18

(1.1092) 1.39

(0.3827) Unemployment Good news -2.19

(1.0023) 0.69

(1.0871) -1.34

(0.4572) Bad news -0.23

(1.6342)* 1.62

(0.0928) 0.88

(1.7261)* Other news Good news -2.03

(2.3928)** 1.38

(0.6251) -1.03

(1.0091) Bad news -1.26

(3.2871)*** -0.98

(1.6211)* -2.46

(1.7654)* Aggregate variable Good news 1.82

(1.1726) 0.93

(1.6421)* 1.09

(0.9876) Bad news -0.76

(1.7652)* 1.28

(0.0982) 1.42

(1.0028) Hungary

CPI Good news -0.72

(0.1827) 1.12

(0.9876) 0.89

(1.1871) Bad news -0.85

(2.1012)** 0.87

(1.0098) 1.21

(0.8765)

Current account Good news -0.21

(0.0561) 1.41

(0.0765) 0.56

(1.1557) Bad news -0.52

(1.0456) -1.11

(1.0074) -1.65

(1.1029) GDP Good news -0.03

(0.0087) 0.96

(2.0125)* 1.02

(0.0987) Bad news -0.73

(2.4612)** -1.51

(1.9864)* -1.26

(1.7074)* Short term interest rate Good news -0.21

(1.1121) 0.67

(1.9987)* 2.07

(2.0198)* Bad news -0.91

(1.9987)** -0.49

(1.1876) -0.97

(0.0076) Terms of trade Good news -0.65

(1.0076) 0.91

(2.6316)** 1.87

(2.6392)** Bad news -0.43

(2.4938)** -0.08

(2.5698)** -0.38

(2.6948)** Unemployment Good news -0.05

(0.0086) 1.01

(1.0098) 0.97

(1.0076) Bad news -0.19

(0.0765) -0.87

(1.6928)* -0.03

(0.1102) Other news Good news -0.11

(0.0928) 1.26

(1.8765)* 0.63

(1.8764)* Bad news -0.23

(1.7615)* -0.07

(2.8726)*** -0.08

(1.8725)* Aggregate variable Good news -0.83

(0.0062) 0.51

(0.9876) 0.15

(0.7181) Bad news -0.19

(1.0081) -0.12

(1.0716) -0.12

(1.0098) Poland

CPI Good news -0.08

(1.1211) 0.41

(1.0928) 0.64

(1.0026) Bad news -0.02

(0.0987) -0.01

(1.0261) -0.04

(1.2191) Current account Good news -0.01

(1.0087) 0.26

(1.7819)* 0.33

(2.3921)** Bad news -0.07

(1.5671)* -0.12

(1.1009) -0.42

(1.1192) GDP Good news -0.03

(0.0987) 0.23

(1.0928) 0.45

(2.0981)* Bad news -0.02

(1.0821) 0.02

(1.0003) -0.52

(1.8171)* Short term interest rate Good news -0.04

(1.0028)* 0.21

(0.0875) 0.66

(0.0987) Bad news -0.01

(1.0612) -0.07

(0.0306) -0.52

(1.8171)* Terms of trade Good news -0.02

(0.0519) 0.64

(1.7651)* 0.95

(2.2121)** Bad news -0.03

(1.0098) -0.34

(2.6718)** -0.71

(1.9182)* Unemployment Good news -0.34

(0.0098) 0.16

(1.9875)* 1.12

(1.8765)*

42

Bad news -0.23 (1.0765)

-0.19 (0.9875)

-0.26 (0.9811)

Other news Good news -0.17

(2.4819)** 0.65

(2.4827)** 1.32

(4.2918)*** Bad news -0.36

(1.9271)* -0.32

(0.08765) -1.12

(0.0876) Aggregate variable Good news -0.02

(0.0098) 0.87

(0.4526) 0.98

(0.2817) Bad news -0.12

(1.0211) -0.63

(0.1827) -0.82

(0.1291) EU

CPI Good news -1.21

(0.1821) 2.19

(1.2011) 0.87

(0.0192) Bad news -0.09

(1.0987) -1.12

(1.0716) -0.11

(2.0291)* Current account Good news -0.91

(0.0181) 2.34

(1.8425)* 1.22

(3.0615)*** Bad news -1.42

(1.0516) -1.87

(1.9981)** -0.24

(1.0182) GDP Good news -1.03

(0.6271) 1.52

(1.0331) 1.09

(1.0082) Bad news -0.67

(1.1921) -1.20

(1.9616)** -0.18

(0.7651) Short term interest rate Good news -0.87

(1.7221)* 0.97

(0.3721) 0.77

(1.1211) Bad news -0.33

(1.0524) -1.06

(2.0871)** -1.83

(1.2821)* Terms of trade Good news -1.61

(2.0987)* 0.90

(6.9211)*** 0.93

(3.2911)*** Bad news -0.11

(0.8711) -1.98

(2.4411)** -2.34

(1.6651)* Unemployment Good news -2.1

(1.0621) 1.16

(1.9987)** 0.88

(2.2811)** Bad news -0.51

(0.4456) -1.02

(1.8876)* -2.11

(2.4131)** Aggregate variable Good news -0.93

(0.9871) 1.76

(1.0917) 1.03

(1.8761)* Bad news -0.01

(0.0987) -0.02

(0.9981) -2.66

(2.5571)** UK

CPI Good news -1.11

(1.7881)* 0.97

(2.0331)** -0.03

(1.5451)* Bad news -0.09

(0.0011) -0.31

(0.6541) -1.13

(2.3133)** Current account Good news -0.98

(1.0541) 1.07

(1.7611)* -0.22

(0.5462) Bad news -0.13

(0.6511) -0.87

(1.6991)* -1.31

(0.1614) GDP Good news -0.76

(4.0611)*** 0.94

(0.9876) -1.50

(0.5875) Bad news -0.29

(1.8761)* -0.66

(1.0987) -1.99

(0.5511)

Short term interest rate Good news -0.61

(1.0654) 0.71

(0.0431) 0.48

(1.7610)* Bad news -0.40

(1.9871)* -0.01

(1.9916)** -0.79

(0.6610) Terms of trade Good news -0.08

(0.0717) 0.71

(0.0391) -1.33

(0.0119) Bad news -0.66

(2.0291)* -1.29

(1.9908)** -2.61

(2.1018)** Unemployment Good news -0.02

(2.1192)* 0.31

(0.3918) -2.01

(0.8711) Bad news -0.27

(1.9562)* -1.44

(0.1102) -2.19

(0.9871) Aggregate variable Good news -1.38

(0.7719) 1.91

(1.1091) 3.07

(4.2711)*** Bad news -0.66

(1.8192)* -1.83

(2.5519)** -1.81

(0.0161)

Adjusted R2 0.068 0.094 0.089 Number of Obs. 640 640 640

Panel-c: Post crisis (1997-end 2004) Constant 1.43

(0.4645) 3.67

(0.3297) 22.00

(0.8751) Czech Republic

CPI Good news 0.18

(0.7048) 0.79

(1.6838)* 1.64

(1.0053) Bad news -0.14

(1.0217) -0.19

(0.6055) -0.71

(0.7086) Current account Good news 0.32

(1.3727) 0.81

(1.7212)* 2.67

(1.5302) Bad news 0.05

(0.3283) 0.10

(0.2647) -0.02

(0.0207) GDP Good news 0.10

(0.4673) 0.07

(0.1617) 0.29

(0.2011) Bad news -0.56

(1.6041) -0.49

(0.6851) -3.43

(1.7108)* Short term interest rate Good news -0.26

(1.3204) -0.50

(1.1396) -1.61

(0.8593) Bad news -0.30

(2.0324)** -0.44

(1.3797) -2.28

(2.0751)** Terms of trade Good news -0.06

(0.4935) -0.20

(0.8172) -0.84

(1.1061) Bad news 0.14

(0.9206) 0.28

(0.7289) 1.58

(1.5003) Unemployment Good news 0.30

(1.669)* 0.63

(1.5771) 2.05

(1.7150)* Bad news -0.11

(0.7722) -0.21

(0.6718) -1.01

(1.0285) Other news Good news -0.15

(1.1425) -0.56

(1.9576)** -2.00

(2.1384)** Bad news -0.25

(1.9459)* -0.53

(1.8009)* -2.27

(2.3432)**

44

Aggregate variable Good news -0.01

(0.0419) -0.06

(0.2207) 0.02

(0.0209) Bad news 0.06

(0.4682) -0.08

(0.2782) 0.42

(0.4681) Hungary

CPI Good news 0.23

(1.1203) 0.62

(1.3280) 0.58

(0.4864) Bad news -0.31

(1.6742)* -0.66

(1.7155) -1.87

(1.5386) Current account Good news 0.04

(0.1946) -0.02

(0.0506) 0.26

(0.1943) Bad news 0.13

(0.4070) 0.33

(0.5787) 1.33

(0.6443) GDP Good news 0.10

(0.6719) 0.28

(0.8515) 0.54

(0.5025) Bad news -0.86

(5.2823)*** -1.90

(5.6227)*** -7.11

(7.0234)*** Short term interest rate Good news 0.47

(3.8805)*** 1.44

(5.6540)*** 5.32

(3.1633)*** Bad news -0.23

(1.1190) -0.66

(1.4995) -0.86

(0.6357) Terms of trade Good news 1.51

(8.5749)*** 2.99

(6.7924)*** 9.51

(8.9892)*** Bad news 1.69

(11.3973)*** 3.74

(15.6235)*** 8.26

(5.2610)*** Unemployment Good news -0.02

(0.2228) -0.11

(0.5413) -0.30

(0.4368) Bad news 0.10

(0.7112) 0.45

(1.7815)* 1.14

(1.1925) Other news Good news 0.10

(1.7338)* 0.35

(2.2576)** 0.69

(1.8892)** Bad news -0.45

(4.5293)*** -1.26

(4.4433)*** -3.46

(4.4992)*** Aggregate variable Good news 0.03

(1.5460) 0.03

(2.4164)** 0.35

(1.8703)* Bad news 0.25

(0.4474) 0.82

(0.3709) 1.90

(0.0930) Poland

CPI Good news 0.29

(1.8691)* 0.74

(2.2809)** 2.64

(2.4852)** Bad news -0.01

(0.0820) -0.14

(0.4983) 0.12

(0.1174) Current account Good news 0.13

(0.6382) 0.60

(1.5682) 1.11

(0.9348) Bad news -0.33

(2.2553)** -0.39

(1.3652) -1.44

(1.3296) GDP Good news -0.11

(0.8576) -0.40

(1.2828) -1.05

(1.0634) Bad news 0.09

(0.2880) 0.57

(0.9009) 2.38

(1.1236) Short term interest rate Good news -0.48

(4.4608)*** -0.42

(1.5353) -2.21

(1.9204)*

Bad news 0.13 (0.8689)

0.23 (0.7336)

2.51 (2.0748)**

Terms of trade Good news -0.19

(1.6815)* -0.56

(1.6919)* -1.48

(1.3692) Bad news 0.07

(0.4152) -0.32

(0.7953) 1.14

(0.9992) Unemployment Good news 0.00

(0.0146) -0.34

(0.8930) 0.46

(0.3396) Bad news -0.02

(0.1560) 0.07

(0.2116) 0.54

(0.4889) Other news Good news -0.35

(4.5293)*** -0.66

(4.4433)*** -2.37

(4.4992)*** Bad news -0.40

(5.0371)*** -0.42

(2.4904)** -1.93

(3.0737)*** Aggregate variable Good news -0.03

(0.4474) 0.06

(0.3709) 0.05

(0.0930) Bad news -0.17

(2.0010)** -0.11

(0.6472) -1.08

(1.6640)* EU

CPI Good news -0.21

(1.0571) -0.18

(0.3998) -1.77

(1.1728) Bad news -0.15

(0.4777) -0.73

(1.5035) -1.23

(0.6169) Current account Good news -0.42

(3.3012)*** -0.90

(2.9325)*** -2.93

(2.5240)** Bad news -0.37

(1.0764) -1.48

(2.8832)** -3.45

(1.6025) GDP Good news -0.33

(2.4890)** -0.63

(1.8538)* -2.52

(2.2747)** Bad news 0.60

(1.4460) 0.32

(0.3890) 3.56

(1.4552) Short term interest rate Good news -0.41

(3.6964)*** -0.32

(0.9517) -2.54

(2.5136)** Bad news 0.05

(0.1256) -0.38

(0.5357) -0.77

(0.2885) Terms of trade Good news -0.33

(2.8495)*** -0.67

(2.2766)** -3.10

(2.9667)*** Bad news -0.40

(1.3854) -1.09

(2.6659)** -2.45

(1.3493) Unemployment Good news -0.70

(4.7857)*** -1.30

(3.6447)*** -4.77

(4.0279)*** Bad news 0.24

(0.6416) -0.62

(0.9225) -0.27

(0.1109) Aggregate variable Good news 0.29

(2.0401)** 0.47

(1.3548) 2.31

(1.9828)** Bad news 0.16

(0.5002) 0.93

(1.8461)* 1.53

(0.7418) UK

CPI Good news -0.75

(3.6752)*** -1.09

(2.5813)** -4.34

(2.5538)** Bad news -0.17

(0.5639) -0.70

(1.4589) -0.97

(0.6943)

46

Current account Good news -0.31

(1.6271) -0.65

(1.8572)* -1.42

(1.2012) Bad news -0.32

(0.9032) -1.19

(2.0248)** -2.74

(1.3867) GDP Good news -0.77

(4.3059)*** -0.67

(1.6789)* -4.16

(2.6599)** Bad news -0.42

(1.9536)* -0.46

(0.9141) -1.50

(0.8647) Short term interest rate Good news -0.91

(2.6856)*** -2.05

(2.4302)** -7.43

(2.8687)*** Bad news -0.65

(2.0577)** -1.21

(2.3506)** -3.72

(2.3878)** Terms of trade Good news -0.85

(5.0102)*** -0.76

(2.0587)** -4.07

(2.7425)*** Bad news 0.03

(0.1033) -0.10

(0.2081) 0.57

(0.4364) Unemployment Good news -1.03

(4.7616)*** -1.37

(2.9843)*** -5.38

(3.0515)*** Bad news -0.01

(0.0371) -0.23

(0.3623) -0.20

(0.1090) Aggregate variable Good news 0.89

(4.5382)*** 1.14

(2.6968)** 4.95

(3.0160)*** Bad news 0.24

(0.7716) 0.90

(1.8428)* 1.76

(1.2336)

Adjusted R2 0.073 0.091 0.094 Number of Obs. 2088 2088 2088

Note: *, ** and *** denote rejection at the 10%, 5% and 1% levels respectively. Absolute value of t-statistics in parenthesis. National Bank Index set as dependant variable.

Table XIII: Regional Shocks REGION DATE COUNTRY TYPE NEWS International 9/4/1998

EU

UK

US

+

+

+

-Tietmeyer sees 'no need' to cut

German interest rates.

-Euro zone growth to be driven by

domestic demand: Trichet.

- Chief executive of British Aerospace

PLC, welcomed moves towards a

single European currency.

- The Federal Reserve Bank of New

York added reserves to the banking

system through a round of four-day

system repos.

International 9/13/2001

EU

UK

US

-

-

-

- War on the world: the silence;

millions to pay respects on world-wide

day of mourning.

Eastern

Europe

10/18/2004

CZ

HU

PL

+

+

+

-State will receive billions of Czech

Crowns from dividends.

-President Klaus signs three new

economic laws today.

-Hungary central bank cuts key

interest.

- Forint strengthens on rate cut.

-Top bank PKO BP sells more shares

to first-time individual investors

in IPO.

-Rapid growth continues to drive

Polish fiscal consolidation.

Eastern

Europe

10/28/2004

CZ

HU

PL

+

+

+

-Central bank governor protests

planned changes to bank law.

-Forint firms on interbank market.

- Central Bank holds Polish interest

rates steady.

48

Table XIV: Modeling Volatility CZ HU PL

constant 0.00 (0.1417)

-0.08 (0.3172)

-0.06 (1.1038)

local shocks CZ 0.00

(1.1883) 0.00

(0.5589) 0.00

(0.1478) HU 0.00

(0.5088) 0.00

(0.6125) 0.00

(0.7725) PL 0.00

(1.0691) -0.01

(2.2367)** 0.00

(0.5908) EU 0.00

(0.8594) -0.01

(2.0484)** 0.00

(1.2383) UK 0.00

(0.1380) 0.01

(1.1135) 0.01

(2.0160) US -0.01

(1.6165) -0.01

(3.3350)*** 0.00

(0.9895) regional shocks Eastern Europe (CZ,HU,PL)

0.00 (0.2769)

0.02 (1.3148)

0.00 (0.0105)

International (EU,US,UK)

0.06 (3.4509)***

0.01 (0.6016)

0.01 (0.5564)

Note: *, ** and *** denote rejection at the 10%, 5% and 1% levels respectively. Absolute value of t-statistics in parenthesis. National Bank Index set as dependant variable.