Political Elections and Stock Price Volatility: The Case of Greece. Str., Chios, Greece. Thessaloniki, Greece.

Similar documents
DEPARTMENT OF ECONOMICS AND FINANCE COLLEGE OF BUSINESS AND ECONOMICS UNIVERSITY OF CANTERBURY CHRISTCHURCH, NEW ZEALAND

Monetary policymaking and inflation expectations: The experience of Latin America

Tourism forecasting using conditional volatility models

Asymmetry and Leverage in Conditional Volatility Models*

Vectorautoregressive Model and Cointegration Analysis. Time Series Analysis Dr. Sevtap Kestel 1

Chapter 5. Heterocedastic Models. Introduction to time series (2008) 1

Licenciatura de ADE y Licenciatura conjunta Derecho y ADE. Hoja de ejercicios 2 PARTE A

R t. C t P t. + u t. C t = αp t + βr t + v t. + β + w t

A Specification Test for Linear Dynamic Stochastic General Equilibrium Models

Volatility. Many economic series, and most financial series, display conditional volatility

Bias in Conditional and Unconditional Fixed Effects Logit Estimation: a Correction * Tom Coupé

Mean Reversion of Balance of Payments GEvidence from Sequential Trend Break Unit Root Tests. Abstract

Granger Causality Among Pre-Crisis East Asian Exchange Rates. (Running Title: Granger Causality Among Pre-Crisis East Asian Exchange Rates)

Financial Econometrics Jeffrey R. Russell Midterm Winter 2009 SOLUTIONS

Department of Economics East Carolina University Greenville, NC Phone: Fax:

Diebold, Chapter 7. Francis X. Diebold, Elements of Forecasting, 4th Edition (Mason, Ohio: Cengage Learning, 2006). Chapter 7. Characterizing Cycles

Forecasting optimally

Modeling the Volatility of Shanghai Composite Index

Ready for euro? Empirical study of the actual monetary policy independence in Poland VECM modelling

Solutions to Odd Number Exercises in Chapter 6

A unit root test based on smooth transitions and nonlinear adjustment

Unit Root Time Series. Univariate random walk

A Dynamic Model of Economic Fluctuations

GDP PER CAPITA IN EUROPE: TIME TRENDS AND PERSISTENCE

Cointegration and Implications for Forecasting

Modeling and Forecasting Volatility Autoregressive Conditional Heteroskedasticity Models. Economic Forecasting Anthony Tay Slide 1

How to Deal with Structural Breaks in Practical Cointegration Analysis

Dynamic Econometric Models: Y t = + 0 X t + 1 X t X t k X t-k + e t. A. Autoregressive Model:

Testing for a Single Factor Model in the Multivariate State Space Framework

AN APPLICATION OF THE GARCH-t MODEL ON CENTRAL EUROPEAN STOCK RETURNS

ACE 562 Fall Lecture 5: The Simple Linear Regression Model: Sampling Properties of the Least Squares Estimators. by Professor Scott H.

ESTIMATION OF DYNAMIC PANEL DATA MODELS WHEN REGRESSION COEFFICIENTS AND INDIVIDUAL EFFECTS ARE TIME-VARYING

International Parity Relations between Poland and Germany: A Cointegrated VAR Approach

Robust critical values for unit root tests for series with conditional heteroscedasticity errors: An application of the simple NoVaS transformation

Time Series Test of Nonlinear Convergence and Transitional Dynamics. Terence Tai-Leung Chong

Problem Set 5. Graduate Macro II, Spring 2017 The University of Notre Dame Professor Sims

DEPARTMENT OF STATISTICS

Regression with Time Series Data

Methodology. -ratios are biased and that the appropriate critical values have to be increased by an amount. that depends on the sample size.

Box-Jenkins Modelling of Nigerian Stock Prices Data

STRUCTURAL CHANGE IN TIME SERIES OF THE EXCHANGE RATES BETWEEN YEN-DOLLAR AND YEN-EURO IN

Nonstationarity-Integrated Models. Time Series Analysis Dr. Sevtap Kestel 1

Comparing Means: t-tests for One Sample & Two Related Samples

Introduction D P. r = constant discount rate, g = Gordon Model (1962): constant dividend growth rate.

Robust estimation based on the first- and third-moment restrictions of the power transformation model

Linear Combinations of Volatility Forecasts for the WIG20 and Polish Exchange Rates

US AND LATIN AMERICAN STOCK MARKET LINKAGES

Wednesday, November 7 Handout: Heteroskedasticity

Asymmetry and Leverage in Conditional Volatility Models

A note on spurious regressions between stationary series

GMM - Generalized Method of Moments

LONG MEMORY AT THE LONG-RUN AND THE SEASONAL MONTHLY FREQUENCIES IN THE US MONEY STOCK. Guglielmo Maria Caporale. Brunel University, London

Scholars Journal of Economics, Business and Management e-issn

Econ Autocorrelation. Sanjaya DeSilva

A New Unit Root Test against Asymmetric ESTAR Nonlinearity with Smooth Breaks

Økonomisk Kandidateksamen 2005(II) Econometrics 2. Solution

UNIVERSITY OF OSLO DEPARTMENT OF ECONOMICS

Yong Jiang, Zhongbao Zhou School of Business Administration, Hunan University, Changsha , China

Asymmetric Conditional Volatility on the Romanian Stock Market - DISSERTATION PAPER-

ECON 482 / WH Hong Time Series Data Analysis 1. The Nature of Time Series Data. Example of time series data (inflation and unemployment rates)

ACE 564 Spring Lecture 7. Extensions of The Multiple Regression Model: Dummy Independent Variables. by Professor Scott H.

Stationary Time Series

Chapter 16. Regression with Time Series Data

ACE 562 Fall Lecture 4: Simple Linear Regression Model: Specification and Estimation. by Professor Scott H. Irwin

ACE 562 Fall Lecture 8: The Simple Linear Regression Model: R 2, Reporting the Results and Prediction. by Professor Scott H.

An Empirical Analysis of the Exchange Rate Volatility: Application of Brazilian and Australian Exchange Markets

What Ties Return Volatilities to Price Valuations and Fundamentals? On-Line Appendix

E β t log (C t ) + M t M t 1. = Y t + B t 1 P t. B t 0 (3) v t = P tc t M t Question 1. Find the FOC s for an optimum in the agent s problem.

Distribution of Estimates

THE IMPACT OF MISDIAGNOSING A STRUCTURAL BREAK ON STANDARD UNIT ROOT TESTS: MONTE CARLO RESULTS FOR SMALL SAMPLE SIZE AND POWER

Macroeconomic Theory Ph.D. Qualifying Examination Fall 2005 ANSWER EACH PART IN A SEPARATE BLUE BOOK. PART ONE: ANSWER IN BOOK 1 WEIGHT 1/3

Dynamic relationship between stock return, trading volume, and volatility in the Stock Exchange of Thailand: does the US subprime crisis matter?

Choice of Spectral Density Estimator in Ng-Perron Test: A Comparative Analysis

Time series Decomposition method

Recursive Modelling of Symmetric and Asymmetric Volatility in the Presence of Extreme Observations *

Currency Depreciation and Korean Stock Market Performance during the Asian Financial Crisis

20. Applications of the Genetic-Drift Model

MODELING AND FORECASTING EXCHANGE RATE DYNAMICS IN PAKISTAN USING ARCH FAMILY OF MODELS

Properties of Autocorrelated Processes Economics 30331

Nature Neuroscience: doi: /nn Supplementary Figure 1. Spike-count autocorrelations in time.

Financial crises and stock market volatility transmission: evidence from Australia, Singapore, the UK, and the US

Econ107 Applied Econometrics Topic 7: Multicollinearity (Studenmund, Chapter 8)

13.3 Term structure models

Estimation Uncertainty

A Hybrid Model for Improving. Malaysian Gold Forecast Accuracy

On Measuring Pro-Poor Growth. 1. On Various Ways of Measuring Pro-Poor Growth: A Short Review of the Literature

14 Autoregressive Moving Average Models

Stock Prices and Dividends in Taiwan's Stock Market: Evidence Based on Time-Varying Present Value Model. Abstract

The Real Exchange Rate, Real Interest Rates, and the Risk Premium. Charles Engel University of Wisconsin

Analysis of Volatility and Forecasting General Index of Dhaka Stock Exchange

Testing for a unit root in a process exhibiting a structural break in the presence of GARCH errors

Variance Bounds Tests for the Hypothesis of Efficient Stock Market

1. Diagnostic (Misspeci cation) Tests: Testing the Assumptions

23.2. Representing Periodic Functions by Fourier Series. Introduction. Prerequisites. Learning Outcomes

Nonlinearity Test on Time Series Data

MODELLING AND FORECASTING DAILY RETURNS VOLATILITY OF NIGERIAN BANKS STOCKS

(10) (a) Derive and plot the spectrum of y. Discuss how the seasonality in the process is evident in spectrum.

Forward guidance. Fed funds target during /15/2017

INVESTIGATING THE WEAK FORM EFFICIENCY OF AN EMERGING MARKET USING PARAMETRIC TESTS: EVIDENCE FROM KARACHI STOCK MARKET OF PAKISTAN

Kriging Models Predicting Atrazine Concentrations in Surface Water Draining Agricultural Watersheds

Transcription:

Poliical Elecions and Sock Price Volailiy: The Case of Greece Ahanasios Koulakiois a, Aposolos Dasilas b and Konsaninos Tolikas c a Universiy of he Aegean, Deparmen of Financial and Managemen Engineering, 3 Foini Sr., 8 Chios, Greece. b Universiy of Macedonia, Deparmen of Accouning and Finance, 56 Egnaia Sr., 546 Thessaloniki, Greece. c Cardiff Universiy, Cardiff Business School, Aberconway Building, Column Drive, Cardiff, CF 3EZ, UK Absrac In his paper he impac of he poliical elecions in Greece on he reurn and volailiy of he Ahens Sock Exchange (ASE) is invesigaed using boh he sandard even sudy mehodology and various univariae GARCH models. The empirical resuls revealed posiive pre- and pos-elecion abnormal reurns bu negaive abnormal reurns on he firs day afer he official announcemen of he elecion resuls. In addiion, he impac of elecion resuls on he ASE reurn and volailiy was invesigaed using GARCH models and i was found ha hese are significanly affeced by he ransiion of he ruling pary. These findings raise doubs for he efficiency of he Greek sock marke and migh have imporan implicaions for invesors wih respec o decisions regarding enering or exiing he marke or invesmen sraegies around ime periods where poliical elecions are going o ake place. Keywords: Poliical elecions, sock price volailiy, Ahens sock exchange, GARCH models Correspondence address: Aposolos Dasilas, Universiy of Macedonia, Deparmen of Accouning and Finance, 46 Egnaia Sr., 546, Thessaloniki, Greece, Tel:++3-394-737, Email: dasilas@gmail.com, - -

. Inroducion Greece is known as he birhplace of democracy and has a long hisory of poliical elecions. In recen years, afer he collapse of he miliary juna in 974 and he resoraion of he parliamenary democracy, he poliical environmen in Greece is sable wih wo poliical paries dominaing he poliical life; he conservaive pary known as New Democracy (ND) and he socialis pary known as he Pan-Hellenic Socialis Movemen (PASOK). These wo poliical paries succeeded each oher in he cabine for he las 3 years wih he socialis pary saying considerably longer in power han he conservaive pary. This paper invesigaes he effecs of poliical changes in he Greek cabine on he behaviour of he Ahens Sock Exchange (ASE). In paricular, he sandard even sudy mehodology described by Dodd and Warner (983) and Brown and Warner (985) is adoped o examine he behavior of he ASE composie index reurn around he elecion daes, while he modified E-GARCH, GARCH and GJR-GARCH models, are employed, as proposed by Lin and Wang (5), o examine he impac of governmen change on sock reurn and volailiy on ASE. Alhough for mos of is hisory he ASE was regarded as a developing sock marke, from he middle of 98s he ASE sared o develop. The driving forces behind his developmen were he Invesmen Services Direcive (EC, 993) aiming a liberalizing he ASE and harmonizing i wih he oher European sock markes, he convergence of he Greek economy o he European requiremens, he sable poliical environmen and improvemens in he echnical infrasrucure. As a resul capial inflows from boh domesic and foreign invesors increased and he ASE developed considerably in erms of marke capializaion, urnover and number of lised companies. During his ime period Greece experienced a raher large number of poliical elecions and herefore, i would be ineresing o examine wheher hese changes had a significan effec on boh he ASE reurn and volailiy. Besides adding o he raher limied lieraure, he resuls of his sudy can also be of paricular imporance o

invesors concerned wih decisions regarding enry o or exi from he marke and changes of invesmen sraegies. The remainder of he paper is organized as follows. Secion briefly reviews he lieraure while secion 3 describes he daa and he mehodology. Secion 4 conains and analyses he empirical resuls and finally secion 5 summarises and concludes he paper.. Lieraure Review The firs probably researcher who analyzed he relaionship beween economics and poliics was Nordhaus (975) who showed ha here is a significan elecion induced economic cycle in he US. I has long been argued ha major poliical evens such as elecions can have a significan impac on he sock marke. For example, Panzalis e al. () found ha sock marke prices end o respond o new informaion regarding poliical decisions ha may affec a naion s fiscal and moneary policy. Oher sudies empirically invesigaed he effecs of economic evens on presidenial voing and he impac of differen poliical srucures o various economic variables (Aesoglou and Congleon, 98; Burdekin, 988). Brasiois (), for example, examined he inflaionary consequences of eleced poliical paries in Greece before and afer is commimen o he Single European Ac (SEA) in 986 and found ha inflaion plays a significan role in he poliical parisan cycle in Greece afer he inroducion of SEA. Anoher se of sudies examined sock marke efficiency around poliical elecion daes. Gemmill (99), for example, found evidence of gross inefficiency in opions prices in UK during he las week of he elecions period implying a decreasing probabiliy of a conservaive pary win while he opinion polls showed he opposie. A number of sudies have also sudied he impac of poliical elecions on sock marke reurns (see, for example, Huang, 985 and The Single European Ac can be considered as he firs formal aemp owards he economic and poliical convergence and inegraion of EU counry members. 3

Foerser and Schmiz, 997). In general, he resuls of hese sudies suppored he so-called presidenial elecion cycle according o which he US sock marke have higher reurns in years 3 and 4 han in years and of a presidenial erm 3. In addiion, Panzalis e al. () invesigaed he sock marke performance around poliical elecions using daa from 33 counries around he world. They found a posiive sock marke reacion in he wo weekperiod preceding elecion daes. This posiive abnormal reurn was sronger for elecions wih higher degree of uncerainy (Similar findings were repored in he lieraure for he case of he UK sock marke by Peel and Pope (983)). Kim and Mei () found ha poliical developmens in Hong Kong had a significan impac on volailiy and reurn while Chan and Wei (996) and Bilingmayer (998) found evidence ha posiive poliical news posiively affec currency and equiy markes. More recenly, Siokis and Kapopoulos (7) examined wheher movemens in he sock prices on he ASE could be parially explained by he dynamics of he poliical environmen. Using an EGARCH-M model and daily daa for he ASE composie index from January 987 o June 4, hey found ha poliical changes indeed impaced he condiional variances on ASE and hey unveiled evidence ha he behaviour of he reurn offered by ASE was affeced asymmerically by pas innovaions. They also repored ha volailiy increases more in he pre-elecion period and when he righ-wing pary is in power. 3. Daa and Mehodology 3. Daa descripion Daily reurns of he ASE composie index were colleced from he ASE Disseminaion Informaion Deparmen for he year period January 985 o December 5. During his period, 7 poliical elecions ook place in Greece; 8/6/89, 5//89, 8/4/9, //93, /9/96, 3 This is mainly because he firs and second year of he presidenial erm are considered o be more appropriae o inroduce unpopular changes such as ax increases. As business profis suffer he negaive effecs of hese policies, earnings shorfalls lead o negaive or low sock marke reurns. 4

9/4/ and 7/3/4. However, wo elecion daes, 5//89 and 8/4/9, were excluded from he sample because elecions ook place again in a shor ime period since he winner pary of 8 h June of 989 could no form a majoriy governmen. In addiion, since elecions in Greece ake place on Sundays, he day afer he elecions was aken o be he day on which he effec on he sock exchange migh be observed. Table repors descripive saisics of he daily reurns of he ASE composie index. The daily mean reurn was.9% and he daily sandard deviaion was.79%. An examinaion of Table also reveals ha he hypohesis ha daily reurns follow a normal disribuion can be rejeced due o he large value of kurosis (3.9); also confirmed by he more formal es of Kolmogorov-Smirnov. The Ljung-Box (LB) es saisic also rejecs he hypohesis ha all auocorrelaions up o lags are zero for boh he reurns and squared reurns which jusifies he use of ARCH-ype models for he variance. [Inser Table abou here] However, even hough he LB saisic provides evidence for second-momen ime dependencies, i canno be used o es he asymmeric reurn volailiy of bad and good news because i is a saisical es which accouns for only he amoun of serial correlaion in he reurn series. Therefore, o invesigae wheher he shocks on he ASE reurn have an asymmeric effec on volailiy, he diagnosics proposed by Engle and Ng (993) are used. These include he (i), sign bias es, (ii) negaive size bias es, (iii) posiive size bias es, and (iv) join es. The firs es examines he impac of posiive and negaive innovaions on volailiy no prediced by he model. In paricular, he squared residuals are regressed agains a consan and a dummy S ha akes he value of one when ε is negaive and zero oherwise. The impac of large and small negaive innovaions on volailiy is capured by he negaive size bias es. I is based on he regression of he sandardized residuals agains a consan and ε. The calculaed -saisic for S ε is used o es for he biases. The posiive sign bias S 5

es examines possible biases associaed wih large and small posiive innovaions. The sandardized filered residuals are regressed agains a consan and (-S ) ε. Again, he - saisic for (- S ) ε is used o es for he possible biases. Finally, he join es uses he F- es based on a regression ha included all hree variables, i.e. S, S ε and (- S ) ε. The calculaed -saisics as well as he F-saisic of hese regressions are repored in Table. The resuls indicaed significan negaive size bias, significan posiive size bias and a significan join F-es, suggesing he presence of asymmeries in he condiional variance. In addiion, he volailiy of he ASE reurns was found o exhibi condiional heeroscedasiciy. Boh he Augmened Dickey Fuller (ADF) and Phillips-Perron (PP) ess rejeced he null hypohesis ha here is a uni roo in he ASE reurns 4 a any convenional significance level (see Table ). The uncondiional kurosis of he ASE daily reurns repored in Table 3 was 3.53. In addiion, he sandardized residuals had zero mean and uni variance. The esimaed Ljung-Box saisics for 5 and lags did rejec he hypohesis of nonlinear dependence in he normalized residuals bu here was no a rejecion in he squared normalized residuals. This means ha an ARCH ype model could be used o describe he behavior of normalized residuals and he behavior of he squared normalized residuals since he auocorrelaions of 5 and lags for he normalized and he squared normalized residuals were saisically significan. Overall, he evidence suppors he inclusion of condiional heeroskedasic and asymmeric componens in he volailiy specificaion in order o model adequaely he ASE volailiy. [Inser Table abou here] [Inser Table 3 abou here] 4 Siokis and Kapopoulos (7) found ha a uni roo exiss in he level of he ASE index. However, hey used sock prices insead of log reurns. 6

3. Mehodology 3.. The mean adjused reurn model The classical even sudy mehodology described by Dodd and Warner (983) and Brown and Warner (985) was employed o esimae he ASE index reurn reacion around he day of an elecion. We define day zero (=) as he firs day following he announcemen of he elecion resul. In Greece, elecions ake place on Sundays and herefore, day zero is he firs Monday afer he elecion day. Using he mean-adjused reurn model, abnormal reurns 5 ( AR ) of he ASE index around he firs pos-elecion day were calculaed as he difference beween he ex-pos reurn R and he normal reurn R : AR = R R () where he normal reurn for he ASE composie index ( R ) is he mean hisorical reurn over a 5-days period prior o he even period, ha is, from day -6 o day -. The even period is a -day window around he firs pos-elecion day ( =); ha is, from = - o = +. The saisical significance of he mean abnormal reurn was esed using he -es, while he saisical significance of he median abnormal reurn was esed using he Wilcoxon signed rank es 6. 3.. ARCH-ype modeling approach A number of sudies have used he so-called GARCH, E-GARCH of Nelson (99) and he GJR-GARCH model of Glosen e al. (993) (see, for example, Bollerslev (986), Friedman and Sanddorf-Kohle () and Siokis and Kapopoulos (7)). The firs model is 5 The ASE logarihmic reurns were calculaed according o he formula R ( P P ) = ln, where P is he index price on day and P - is he index price on day -. 6 The Wilcoxon signed rank es, also known as he Wilcoxon mached pairs es, is a non-parameric es used o es he median difference in paired daa. This es is he non-parameric equivalen of he paired -es. The Wilcoxon signed rank procedure assumes ha he sample we have is randomly aken from a populaion which has a symmeric probabiliy disribuion. The symmeric assumpion does no assume normaliy; i simply assumes ha here is roughly he same number of values above and below he median. 7

symmeric while he laer wo are asymmeric models. Having assessed he abiliy of all hree models o describe he daily ASE reurns volailiy, he effecs of ransiion of ruling pary on he sock marke behavior was also examined. Dummies were embedded in he hree above menioned models o deec he effec of ransiion of ruling pary as follows: R m + D + D + bi R i ε () i= = + where E( ε ) = µ R, ε Ω ~ T (, h ). D denoes he dummy variable ha akes he value of for he ransiion of ruling pary and oherwise. The second dummy variable D conrols for he sock marke crash of Ocober 987 where here was a large increase in volailiy. Therefore, he sample period is broken ino he pre-987 and pos-987 period. In paricular dummy D equals for he pos-987 period and for he pre-987 period. The symmeric response o shocks is aken from Bollerslev s (986) GARCH model: h τ τ τ β β ε (3) = + + + + D D h he parameer resricions, τ >, β, β and β +β <, ensure ha he sochasic process ( ) Var ( ) = h ε is well-defined (i.e., h > ) and he covariance is saionary wih ( ε ) =, ε and cov (, ) = ε. ε s To allow for asymmeric volailiy effecs he E-GARCH and he GJR-GARCH models were considered. The E-GARCH asymmeric volailiy model is given by: Ε ln h τ τ τ θ β (4) = + D + D + [ u Ε u + u ] + ln h where u ε / = h. The news j ε impac on condiional volailiy ln(h ). When p = q =, he model capures an asymmeric response because ln h lnh / ε / ε = ( θ + ) when ε = ( θ ) when ε - - > <., and volailiy is minimised in he absence of news, ε =. 8

The GJR-GARCH asymmeric volailiy model is described by: h = + τ D + τ D + βh + β ε + β 3 S ε τ (5) where S = if if ε ε e < he process is well-defined when p, q, τ >, i=,,3,..., p, β >, j =,,3,..., q. The lags of he hree models of he condiional mean reurns were chosen as o minimize he value of he Akaike informaion crierion as well as he Schwarz Bayesian Crierion 7. The Maximum Likelihood (ML) esimaion mehod was used o joinly deermine he parameers of he mean and he ime-varying condiional variance-covariance equaions 8. 4. The Empirical Resuls 4. Sock marke reacion around he elecion daes Table 4 repors abnormal reurns (ARs) for he even period sared days before elecion daes (=-) and ended days afer elecion daes (=+). I can be noiced ha he average and median ARs are posiive on day - and equal o.% and.6%, respecively, and saisically significan a he % significance level. On he firs rading day afer he elecion resul becomes known (day ), he ASE reacs negaively having a mean (median) abnormal reurn equal o -.9% (-.57%). The sign of he abnormal reurn becomes posiive on days and, wihou, however, being saisically significan (a mean equal o.44% and.4% on days + and +, respecively). This resul can be aribued o he fac ha he elecion resul is officially announced a he end of he following working day (Monday in our case) and herefore, he sock marke incorporaes ha informaion one day laer. The posiive 7 The Akaike and Schwarz crieria can be used o choose he order of a GARCH model by aking ino accoun boh he model fi and complexiy. 8 The BHHH algorihm proposed by Bernd e al. (974) was used o obain he maximum likelihood esimaes of he parameers. 9

reacion of he ASE before he elecion day can be aribued o he formaion of invesors expecaions ha he new governmen will fulfill is pre-elecoral promises. These resuls are in line wih hose of Panzalis e al. () who also found posiive marke reacion prior o elecion daes and negaive marke reacion on he elecion dae, even hough hey used weekly daa insead of daily ones. [Inser Table 4 abou here] 4.. Sock marke reurns and volailiy around elecion daes Tables 5 o 7 repor he coefficiens of ransiion of ruling pary dummies, (.8E- for GARCH, E-GARCH and GJR-GARCH models) and τ (.76E-3 for he GARCH model,.7 for he E-GARCH model and.76e-3 for he GJR-GARCH model). These esimaes are saisically significanly a he % significan level. Therefore, he ransiion of ruling pary in Greece has an imporan impac on he ASE reurn and volailiy. These finding holds for all of he hree modified models of GARCH, E-GARCH and GJR-GARCH regardless of he impac of news being symmeric or asymmeric. This finding is in conras o ha of Lin and Wang (5) who found no significan relaionship beween he dummy of ransiion of ruling pary and he sock reurns and volailiy of he Nikkei 5 sock index. The values of he dummy variable ha conrols for he 987 sock marke crash (-.8E-3 for he GARCH, E-GARCH and GJR-GARCH models) and τ (.44E-5 for he GARCH model,.5e- for he E-GARCH model and.45e-5 for he GJR-GARCH model) indicaed ha he ASE reurn was negaive and no significan, while he ASE volailiy was significanly posiive a he 5% significance level. Unsurprisingly, he ASE index reurn volailiy, τ, was found o be significan a he % level and posiively relaed o he 987 sock marke crash (similar findings were repored in he lieraure by Schwer (99), Engle and Musafa (99) and Lin and Wang (5)). The hree models also capure he negaive sign of he dummy variable ha conrols for he 987 sock marke crash for he ASE reurns;

however, he resuls were no saisically significan. These resuls, however, are in conras wih he findings of Lin and Wang (5) who found a saisically significan impac of he 987 crash on sock price reurns. This migh be aribued o he differen developmen sages ha he wo sock markes were in 987. The values of he log-likelihood funcion (9747.4, 9748.3, and 9747.54 for he GARCH, E-GARCH and GJR-GARCH models, respecively) do no indicae grea difference in he amoun of volailiy and noise examined by he hree models. Indeed, he amouns of pas volailiy were equal o.77,.95 and.77 for he GARCH, E-GARCH and GJR-GARCH models, respecively. In addiion, he amoun of pas noise for he GARCH and for he GJR- GARCH model was.. There is no analogous coefficien for he case of E-GARCH model as his model capures he impac of bad and good news arising from shocks wih differen signs. In paricular, if ε < (bad news) hen he impac of bad news on volailiy 9 was equal o -.39 = (.38(-.-)) and if ε (good news) hen he impac of good news on volailiy was equal o.37 = (.38(-.+)). The analogous coefficien of he GJR- GARCH model for he case of bad news is equal o -.88E-3. This coefficien indicaes ha he impac of bad news on volailiy was smaller in magniude in he GJR-GARCH model compared o he E-GARCH model. The difference is due o he fac ha he value of was large and saisically significan a he % significance level in he E-GARCH model. Thus, he impac of noise on volailiy has a bigger shor-erm effec wih he E-GARCH model compared o he saisically insignifican impac of bad news on volailiy in he GJR-GARCH model. [Inser Table 5 abou here] [Inser Table 6 abou here] θ θ 9 The formula used is ( ) The formula used is ( +)

[Inser Table 6 abou here] 5. Summary and Conclusion This sudy examined he impac of he Greek poliical elecions over he period 985 o 5 on he ASE reurns around he elecion daes by employing he sandard even sudy mehodology and ARCH-ype models. The empirical resuls indicaed posiive sock marke reacion on he las working day prior o elecion dae and negaive marke reacion on he firs pos-elecion day. However, he sign of he marke reacion became posiive shorly afer he official resul became known and he sock marke absorbed he news. In addiion, he impac of elecion resuls on he ASE reurn and volailiy was invesigaed and i was found ha hese are significanly affeced by he ransiion of he ruling pary. The impac of he Ocober 987 sock markes crash on he ASE reurn (volailiy) was found o be negaive (posiive) and no saisically significan a he 5% significance level (saisically significan a he 5% significance level). The resuls of his paper migh have imporan implicaions for invesors wih an ineres in he Greek sock marke. In paricular, hey can affec decisions regarding he enry or exi of he ASE and he change of invesmen sraegies. In spie of he day of elecions, here was an abnormal posiive reacion before he day of he elecions which was followed by a negaive abnormal reurn on he firs pos-elecion day and a posiive reurn wo days laer. This resul raises doubs for he efficiency of he Greek sock marke since i appears ha he ASE needs some ime o decode he elecion news. The 987 sock marke crash was also found o have a significan impac on he ASE volailiy; however, he impac was insignifican for he sock reurns. This shows ha he Greek sock marke is influenced differenly in he firs momen han he second momen of reurns and, herefore, an ARCH-ype model would be useful o accoun for his feaure.

References Aesoglou, S.H., and Congleon, R. (98) Economic condiions and naional elecions possample forecass of he Kramer equaions, American Poliical Science Review, 76, 873-875 Bilingmayer, G. (998) Oupu, sock volailiy, and poliical uncerainy in a naural experimen: Germany, Journal of Finance, 53, 43-58. Bollerslev, T. (986) Generalized auoregressive condiional heeroskedasiciy, Journal of Economerics, 3, 37-37. Brasiois, G.J. () Poliical paries and inflaion in Greece: The meamorphosis of he socialis pary on he way o EMU, Applied Economics Leers, 7, 45-454. Brown, S. and Warner, J. (985) Using daily sock reurn: The case of even sudies, Journal of Financial Economics, 4, 3-3. Burdekin, R.C.K. (988) Economic performance and he deerminaion of presidenial elecions in he US, American Economis, 3, 7-75. Chan, Y. and Wei, J. (996) Poliical risk and sock price volailiy: The case of Hong Kong, Pacific Basin Finance Journal, 8, 47-53. Dodd, P. and Warner, J. (983) On corporae governance: a sudy of proxy coness, Journal of Financial Economics,, 4-438. Engle, R.F. (98) Auoregressive condiional heeroskedasiciy wih esimaes of he variance of Unied Kingdom inflaion, Economerica, 5, 987-7. Engle, R.F. and Ng, V. (993) Measuring and esing he impac of news on volailiy, Journal of Finance, 45, 749-777. European Commission (993) On Invesmen Services in he Securiies Field, Official Journal of he European Communiies, No L. Foerser, S.R. and Schmiz, J.J. (997) The ransmission of US elecion cycles o inernaional sock reurns, Journal of Inernaional Business Sudies, 8, -7. Gemmill, G. (99) Poliical risk and marke efficiency: Tesed based on Briish sock and opion markes in he 987 elecion, Journal of Banking and Finance, 6, -3. Glosen, L.R., Jagannahan, R., and Runkle, D. (993) On he relaion beween he expeced value and he volailiy of he nominal excess reurn on socks, Journal of Finance, 48, 779-8. Huang, R.D. (985) Common sock reurns and presidenial elecions, Financial Analyss Journal, 4, 58-6. Kim, H.Y. and Mei, J.P. () Wha makes he sock marke jump? An analysis of poliical risk on Hong Kong sock reurns, Journal of Inernaional Money and Finance,, 3-6. 3

Lin, C.Y., and Wang, Y. (5) An analysis of poliical changes on Nikkei 5 sock reurns and volailiies, Annals of Economics and Finance, 6, 69-83. Nelson, D. (99) Condiional heeroskedasiciy in asse reurns: A new approach, Economerica, 59, 347-37. Nordhaus, W.D. (975) The Poliical Business Cycle, Review of Economic Sudies, 43, 69-9. Panzalis, C., Sangeland, D.A., and Turle, H.J. () Poliical elecions and he resoluion of uncerainy: The inernaional evidence, Journal of Banking and Finance, 4, 575-64. Peel, D. and Pope, P. (983) General elecions in he UK in he pos 95 period and he behavior of he sock marke, Invesmen Analys, 67, 4-. Siokis, F. and Kapopoulos, P. (7) Paries, elecions and sock marke volailiy: Evidence from a small open economy, Economics and Poliics, 9, 3-85. 4

Table. Descripive saisics of he ASE composie index daily reurns Mean (µ ).9 Sandard Deviaion (σ ).79 Skewness (S).3* Kurosis (K) 3.9* Kolmogorov-Smirnov es saisic.96* LB () 8.33* LB () 67.38* Noe: This able conains descripive saisics for he ASE daily reurns over he period 985 o 5. * denoes saisical significance a he 5% significance level. µ, σ, S and Κ are he mean, sandard deviaion, skewness and kurosis, respecively. The Kolmogorov-Smirnov saisic ess he hypohesis ha he ASE reurns are normally disribued (he criical value a he 5% level is.36/ n, where n is he sample size). LB() and LB () are he lags Ljung-Box saisics calculaed for he reurns and he squared reurns, respecively. The LB saisic is disribued as x. Table. Volailiy specificaion ess for filered reurns Sign bias (-ess) -.96 Negaive size bias (-ess) -.59* Posiive size bias (-ess).6* Join es F(3,57) 49.55* ARCH (4) 8.76* ADF () -47.5* Phillips-Perron -56.99* Noe: This able repors he ess proposed by Engle and Ng (993). These ess invesigae wheher he reurn shocks on ASE have an asymmeric effec on volailiy and are specified as follows: Sign bias: z = + bs + e Negaive sign bias: Posiive sign bias: z = Join es: = + bs ε e + b ( S ) ε + e = + b S + bsε + b3 ( S ) ε z + z + e Where S is a dummy variable ha akes he value of one if ε is negaive and zero oherwise. All -saisics refer o he coefficien b in he firs hree regressions, while he join es F (3, 57) is referring o he forh regression. The normalized residuals z = ε / σ are based on an AR () model applied o he daily reurns. ARCH denoes he Lagrange muliplier es of Engle (98) and he criical value is 7.8 a he 5% significance level. ADF denoes he Augmened Dickey Fuller es saisic and he lag inerval is deermined by minimizing he AIC and SBC values. The funcions of he AIC and SBC are: AIC (k) = T* ln σ + k SBC(k) = T*lnσ + k*lnt Where k denoes he lagged period, T denoes he number of sample, and σ denoes he lagged T k periods of ε.the criical value for he ADF es is equal o -.863. * Denoes saisical i= significance a he 5% significance level. Boh he ADF and Phillips-Perron ess rejec he null hypohesis ha he ASE reurns has a uni roo a any convenional significance level. 5

Table 3. Diagnosic ess for residuals Kurosis (K) 3.53* E(z ). E(z ).99 LB(5) 34.4* LB() 38.3* LB (5) 5.96* LB ().87* Noe: * Denoes saisical significance a he 5% significance level. z is he model normalized residual. LB(.) and LB (.) are he Ljung-Box es saisics for he z and using 5 and lags respecively. z, Table 4. ASE reurn behaviour over a period of days around elecion days Day Mean % p-value (-es) Median % p-value (Wilcoxon signed rank es) -.6.944 -.3.59-9 -.58.36 -.37.59-8.5.796.8.787-7.48.63.65.78-6.7*.57.5*.59-5 -.8.94.7. -4 -..533 -..48-3 -.55.93 -.3*.59 -.6.45 -.5.787 -.*.69.6*.59 -.9.76 -.57.8.44.496.47.59.4.867 -.6. 3 -.95.385 -.6.48 4 -.54. -..6 5-3.8.44 -.75*.59 6 -.3.883.6. 7.3.6.55.8 8.4.89 -.44.787 9.89.44.5.59 -.8.48 -.86.59 Noe: The mean abnormal reurns (ARs) of he ASE index is he difference beween he expos reurn R and he normal reurn R which is he mean hisorical reurn over a 5-days period prior o he even period, ha is, from day -6 o day -. The Wilcoxon signed rank es is a non-parameric es used o es he median difference in paired daa. * denoes ha an esimae is saisically significan a he % significance level. 6

Table 5. The empirical resuls of AR()-GARCH (,) model R = + D + D + b R + ε E( ε ) = µ R h D D = τ + τ D + τ D denoes he dummy of he change of ruling pary and denoes he dummy of 987 crash. + β h + β ε Variable Reurn Variable Volailiy 3 3 * τ.35e-5*** (.7) (9.438) 8 * τ.76e-3*** (.89) (3.765) -8 4 τ.44e-3*** (-.9) (6.554) b.*** (5.77) Coefficien Esimaion µ.48e-3*** (3.59) β.77*** (.736) β.*** (.76) Loglikelihood 9747.4 Noes: (i) Numbers in parenheses are -saisics. (ii)***, ** and * indicae saisical significance a he %, 5% and % significance levels respecively. (iii) This able shows he impac of he change of ruling pary, he 987 sock marke crash and he previous day s reurn on nex day s reurns. (iv) The able also shows he impac of he above wo menioned dummy variables plus he previous day s volailiy and noise on he nex day s volailiy. The equaion E( ε ) = µ R follows a maringale process, while he oher wo equaions comprise a GARCH model which has wo dummy variables which explain he impac of he change of ruling pary and he 987 sock marke crash on volailiy. (v),,, and b denoe he coefficiens for he consan, he dummy for he change of ruling pary, he dummy for he 987 sock marke crash and he previous day s reurns, respecively. (vi) τ, τ, τ, µ, β, β are he coefficiens for he consan, he dummy for he change of ruling pary, he dummy for he 987 sock marke crash, he long erm consan of he maringale model, he previous day s volailiy and he previous day s noise, respecively. 7

Table 6. The Empirical Resuls of AR()-EGARCH (,) R = + D + D + b R + ε E( ε ) = µ R h D D = τ + τ D + τ D + [ u E u denoes he dummy of he change of ruling pary and denoes he dummy of 987 crash. + θu ] + β ln h Variable Reurn Variable Volailiy 3 3 * τ -.43*** (.7) (-3.39) 8 * τ.7*** (.89) (6.959) -8 4 τ.5e-*** (-.9) (5.639) b.*** (5.77) Coefficien Esimaion µ.35e-3*** (.797) β.95*** (99.365).38*** (3.44) θ -.E- (-.7) Loglikelihood 9748.3 Noes: (i) Numbers in parenheses are -saisics. (ii) ***, ** and * indicae saisical significance a he %, 5% and % significance levels, respecively. (iii) This able shows he impac of he change of ruling pary, he 987 sock marke crash and he previous day s reurns on nex day s reurns. (iv) The able also shows he impac of he above wo menioned dummy variables plus he previous day s volailiy and posiive or negaive noise on he nex day s volailiy. The equaion E( ε ) = µ R follows a maringale process, while he oher wo equaions comprise an E-GARCH model which have wo dummy variables which explain he impac of he change of ruling pary and he 987 sock marke crash on volailiy. (v),,, and b are he coefficiens for he consan, dummy for he change of ruling pary, dummy for he 987 sock marke crash and he previous day s reurns of sock, respecively. (vi) τ, τ, τ, µ, β, andθ are he coefficiens for he consan, he dummy for he change of ruling pary, he dummy for he 987 sock marke crash, he long erm consan of he maringale model, he coefficien of he previous day s logarihmic volailiy, he coefficien of caching he impac of bad and good news (noise) and finally he impac of previous day s noise, respecively. 8

Table 7. The Empirical Resuls of AR()-GJR GARCH (,) R = + D + D + b R + ε = τ where E ( ε ) = µ R h S - + τ D + τ D if ε = { if ε - - < + β h i + β ε + β S 3 ε D denoes he dummy of he change of ruling pary D denoes he dummy of 987 crash. Variable Reurn Volailiy 3 3 * τ.35e-5*** (.7) (9.364) 8 * τ.76e-3*** (.89) (3.763) 4 8 τ.45e-5*** (-.9) (6.6) b.*** (5.77) Coefficien Esimaion µ.5e-3*** (3.44) β.77*** (.86) β.*** (9.9) β -.88E- 3 (-.675) Loglikelihood 9747.54 Noes: (i) Numbers in parenheses are -saisic. (ii) ***, ** and * indicae saisical significance a he %, 5% and % significance levels, respecively. (iii) This able shows he impac of he change of ruling pary, he 987 sock marke crash and he previous day s reurns on nex day s reurns. (v) The able also shows he impac of he above wo menioned dummy variables plus he previous day s volailiy and negaive noise on he nex day s volailiy. The equaion E( ε ) = µ R follows a maringale process, while he oher wo equaions comprise a GJR-GARCH model which have wo dummy variables which explain he impac of he change of ruling pary and he 987 sock marke crash on volailiy. (iv),,, and b are he coefficiens for he consan, dummy for he change of ruling pary, dummy for he 987 sock marke crash and he previous day s reurns of sock, respecively. (v) τ, τ, τ, µ, β, β andβ 3 are he coefficiens for he consan, he dummy for he change of ruling pary, he dummy for he 987 sock marke crash, he long erm consan of he maringale model, he coefficien of he previous day s volailiy, he coefficien of previous day s noise and he coefficien of he previous day s negaive noise, respecively. and 9