PRICE DISCOVERY IN THE ATHENS DERIVATIVES EXCHANGE: EVIDENCE FOR THE FTSE/ASE-20 FUTURES

Transcription

1 PRICE DISCOVERY IN THE ATHENS DERIVATIVES EXCHANGE: EVIDENCE FOR THE FTSE/ASE-20 FUTURES MARKET (a) Dimiris F. Kenourgios, Deparmen of Accouning and Finance, Ahens Universiy of Economics and Business, 5 Sadiou Sr., Office 115, Ahens , Greece. Tel: , fax: , dkenourg@econ.uoa.gr. (a) This paper has been published in Economic and Business Review, Vol. 6, No 3 (Oc. 2004), pp

2 PRICE DISCOVERY IN THE ATHENS DERIVATIVES EXCHANGE: EVIDENCE FOR THE FTSE/ASE-20 FUTURES MARKET Absrac: The FTSE/ASE-20 fuures marke, as he firs organised Greek derivaives marke, esablished in Augus 1999 and is operaion ress wih he Ahens Derivaives Exchange (ADEX) and he Ahens Derivaives Exchange Clearing House (ADECH). Coinegraion ess are used and an error correcion model is developed in order o examine he relaionship beween price movemens of FTSE/ASE-20 hree-monh fuures index and he underlying cash marke in Ahens Sock Exchange (ASE). Τhe invesigaion of is price discovery mechanism has been moivaed by he exising pauciy of similar research in such newly esablished (emerging) fuures markes and he growing imporance of his marke for boh invesors and he Greek capial marke. The resuls show he presence of a bi-direcional causaliy beween sock index spo and fuures markes, indicaing ha he newly esablished ADEX can provide fuures conracs ha serve as a focal poin of informaion assimilaion and fulfil heir price discovery. JEL Classificaion: G13, G14 Key words: Ahens Derivaives Exchange, FTSE/ASE 20 fuures conrac, Price discovery, Coinegraion analysis, Causaliy. 2

3 1. INTRODUCTION The relaionship beween sock index spo and fuures markes is sill aracing he aenion of academics, praciioners and regulaors due o boh he considerable volume of rading in hese conracs and heir role during periods of urbulence in financial markes. An imporan aspec of his relaionship is he naure of he lead-lag relaionship in he reurns beween equivalen asses raded in differen markes or he predicive power of price movemens in one marke for hose in he oher marke. One of he economic funcions of fuures conracs is price discovery. Price discovery refers o he use of fuures prices for pricing cash marke ransacions and is significance depends upon he above menioned, close relaionship beween he prices of fuures conracs and he underlying asses. The essence of he price discovery funcion of fuures markes hinges on wheher new informaion is refleced firs in changes of fuures prices or in changes of cash prices. In oher words, price discovery means wheher price changes in fuures markes lead price changes in cash markes more ofen han he reverse. If ha is he case, here exiss a lead-lag relaionship beween he wo markes. Therefore, he fuures prices may serve as he marke s expecaion of a subsequen delivery period cash price. The share of price discovery originaing in he fuures markes has imporan implicaions for hedgers and arbirageurs who use hese markes. The firs sudies o es he price ransmission process have used mainly he regression analysis. However, if price series are no saionary, a phenomenon ypical in financial markes, hen sandard saisical ess of parameer resricions are no reliable (Elam and Dixon, 1988). Thus, for overcoming he problems of non-saionary price series and due o he fac ha price discovery deals wih shor-run and long-run 3

4 deparures from a presumed equilibrium relaion, he inroducion of coinegraion analysis wih error correcion models is foruious. An overwhelming number of sudies have examined he price discovery process involving well esablished US, European and Asian fuures markes providing differen resuls. Noable sudies using Unied Saes daa and differen economeric echniques (e.g., Ng, 1987; Kawaller, P. Koch and W. Koch, 1987; Soll and Whaley, 1990; Chan, 1992; Anoniou and Garre, 1993; Pizzi, Economopoulos and O Neal, 1998) generally suppor he primacy of fuures in he price discovery process. Inernaional evidence supporing he primacy of fuures is no as srong. For insance, Grübichler, Longsaff and Schwarz (1994) and Booh, So and Tse (1999) repor ha he DAX index lags he price of is fuures conrac, a finding generally echoed by Tang and Ho (1989) for he SIMEX and Iihara, Kao and Tokunaga (1996) for he Japanese marke. However, Shyy, Vijayraghavan and Sco-Quinn (1996) repor he opposie for he French spo and fuures prices. Moreover, Wahab and Lashgari (1993) repor a uni-direcional relaionship among spo and fuures prices of S&P 500 and FTSE-100 index, alhough he dominance of he spo prices has been found o be sronger. The purpose of his paper is he examinaion of he informaion linkage beween he FTSE/ASE-20 sock index and is hree-monh index fuures conrac and he role (lead or lag) ha he fuures marke plays using daily closing fuures and cash prices. This invesigaion is significan for wo reasons. Firs, i is focused in an emerging fuures marke, such as is he case for Greece, given he exising pauciy of research in such markes. The findings of his sudy may sugges if newly esablished fuures markes can also provide fuures conracs ha fulfill heir price discovery funcion. Second, given ha here has been no prior invesigaion in ADEX due o is recen 4

5 creaion and shor rading hisory and he FTSE/ASE-20 fuures conrac has he greaer liquidiy among he oher derivaives producs, his examinaion wih more up-o-dae economeric ess han were employed in he early lieraure on price discovery is cerainly of concern o exising and fuure paricipans. Engle-Granger and Johansen coinegraion ess are used and an error correcion model (ECM) is developed in order o examine he causaliy relaionship beween he wo markes. This paper proceeds as follows: Secion 2 gives a brief discussion of he Ahens Derivaives Exchange and he FTSE/ASE-20 fuures conrac. Secion 3 presens he mehodology followed. Secion 4 repors he sample daa and presens he empirical resuls. Secion 5 draws a summary and he conclusions referring he relaionship beween he FTSE/ASE-20 fuures marke and he underlying cash marke. 2. ATHENS DERIVATIVES EXCHANGE (ADEX) AND THE FTSE/ASE-20 FUTURES CONTRACT Unil he lae years of he las decade, and prior o he creaion of he insiuional framework for he operaion of he organized derivaives marke in Greece, ransacions on derivaives exised on a limied scale, over-he couner, mainly beween financial insiuions and companies. The developmen of he organized derivaives marke in Greece, similarly wih oher developed European counries, was a resul of he growh of he Greek capial marke and economy in general 1. The esablishmen of he Ahens Derivaives Exchange (ADEX) and he Ahens Derivaives Clearing House (ADECH) in accordance wih Law 2533/1997 offers a majoriy of sandardized producs o an 1 The Greek capial marke seemed ready o suppor an organised marke on financial derivaives. The urnover raio is over 30% since 1994 whils, in 1998, i increased 62%. Accordingly, he oal capialisaion has been over 20 billion dollars since 1996, reaching 81 billion dollars in Finally, he raio of capialisaion o GDP has increased coninuously from In 1998 i had doubled compared o i for 1997 and was well above 120% in

6 enlarging number of paricipans (corporaions, individual invesors, banks, muual funds, sae enerprises, invesmen companies), conribues o he efficiency of he capial marke and has posiive influence on he naional economy. ADEX and ADECH were founded in April 1998 as auonomous companies. ADEX s purpose is o organize and suppor rading in he derivaives marke. I is organized along wo main axes. The firs is he developmen of business and he second is relaed o he execuion of ransacions. The purpose of ADECH is o ac as counerpary in all rades concluded on ADEX, he clearing of ransacions ha are effeced, he selemen of he ransacions, he ensuring of he fulfillmen of obligaions arising from hese ransacions, and co-operaion wih members and banks, o ensure he safe commimen and disengagemen of margins, he financial selemen of ransacions and every relaed aciviy. The elecronic sysem provided by ADEX is par of he Inegraed Auomaed Elecronic Trading Sysem (OASIS). All ransacions on sandardized derivaives are effeced hrough his sysem, creaing an elecronic marke in which access is via a compuer insallaion a every member s locaion. Direc access o ADEX and ADECH is resriced o hose organizaions, which have been acceped as members, having fulfilled he legal requiremens and submied he deails required by he membership applicaion. There are wo ypes of membership in ADEX. The firs caegory is he single members who ac as broker-agens and are no allowed conducing ransacions for heir own accoun and he second caegory is he marke makers. The FTSE/ASE-20 blue chip index was he firs underlying asse of he fuures conrac, followed by he FTSE/ASE-40 midcap index, he en year Greek bond, he hree-monh ATHIBOR, and seleced blue-chip socks much laer, while American 6

7 syle opions conracs on major Greek blue-chip socks and he above FTSE indexes and sock lending conracs have been launched recenly. The rading on ADEX began on 27/8/1999 and he firs raded produc was he FTSE/ASE-20 fuures conrac. The FTSE/ASE 20 index has been chosen as he mos suiable due o he high liquidiy, and urnover of is consiuen shares. The fuures on he FTSE/ASE-20 are cash seled and quoed in index poins. A any poin in ime, here are six index fuures conracs lised, corresponding o he associaed expiraion monhs: he hree neares consecuive monhs from he monhly cycle and he hree neares monhs from he March, June, Sepember and December quarer cycle, no included in he consecuive monhs. The expiraion day and he las rading day on he FTSE/ASE-20 fuures is he hird Friday of he expiraion monh. Open posiions on fuures are subjec o daily selemen (marking o marke). Table 1 displays he main specificaions of he FTSE/ASE-20 fuures conrac. 7

8 TABLE 1: Specificaions of he FTSE/ASE- 20 Fuures Conrac PRODUCT SETTLEMENT MINIMUM LOT SIZE CONTRACT SIZE QUOTE UNIT MINIMUM TICK TICK VALUE PRICE LIMIT TRADING HOURS MARGIN REQUIREMENTS MARGINING SYSTEM POSITION LIMITS LAST TRADING DATE SETTLEMENT DATE - FTSE/ASE-20 INDEX FUTURES - Cash selemen - Single Marke: 1, Block Marke: EURO per index poin - Index poins index poins - 1,25 EURO - No price limi - Monday o Friday: 10:45 am o 16:15 pm (local ime) - 12% of he posiion - RIVA (Risk Valuaion) per end clien - No posiion limis - 3rd Friday of he expiraion monh - Firs working day following he las rading day LISTING RULES - 3 closes consecuive monhs plus 3 closes from he Mar-Jun-Sep- Dec quarer cycle. On he working day following he las rading day, a new series is inroduced SETTLEMENT OF FEES - Fees are seled on he working day ha follows he rade day (T+1) EXCHANGE FEE - 0,15-0,55 EURO (Marke Makers B) / 1,30-1,80 EURO (Ohers Members) MARGIN - Collaeral using RIVA (Risk Valuaion) a end clien level by Clearing House Source: ADEX 8

9 3. METHODOLOGICAL ISSUES 3.1 Saionariy The exisence of uni roos is firsly esed using he Augmened Dickey-Fuller es (ADF) (Dickey and Fuller, 1981) hrough he following relaionship: k 1 + γ i i= 1 S = α + βτ + ρs S + u (1) i where S = S S 1, S is he index of he spo marke, and k is chosen so ha he deviaions u o be whie noise. The same relaionship is used o deermine he order of he fuures price index ( F ). The null and he alernaive hypohesis for he exisence of uni roo in S and F is H o : ρ = 0, H 1 : ρ 0. If he null hypohesis of only a uni roo canno be rejeced, hen he sock prices follow a random walk. Phillips and Perron (1988) have modified he ADF es (based on Equaion 1 wihou lagged differences), as he ADF ess are only valid under he crucial assumpion of i.i.d. processes. In pracice, i may be more realisic o allow for some dependence among he u s. In ha case, he asympoic disribuion is changed. Phillips and Perron (1988) have weakened he i.i.d. assumpion by using a non-parameric correcion o allow for some serial correlaion and heeroskedasiciy: y = α 0 + a y -1 + u (2) The PP es ends o be more robus o a wide range of serial correlaions and ime-dependen heeroskedasiciy. In he PP es, he null hypohesis is ha a series is non-saionary (i.e. difference saionary) if α = 1, hence, rejecion of he uni roo hypohesis is necessary o suppor saionariy. The asympoic disribuion of he PP - saisic is he same as he ADF -saisic. 9

10 3.2 Coinegraion Evidence of price changes in one marke generaing price changes in he oher marke so as o bring abou a long-run equilibrium relaionship is given in eq. (3): F δ δ1 S = ε (3) o where F and S are conemporaneous fuures and cash prices a ime ; δ 1 andδ o are parameers; and ε is he deviaion from pariy. If F and/or S are nonsaionary hen he Ordinary Leas Square (OLS) mehod is inappropriae because he sandard errors are no consisen. This inconsisency does no allow hypohesis esing of he coinegraing parameerδ 1. If F and S are nonsaionary bu he deviaions, ε, are saionary, F and S are coinegraed and an equilibrium relaionship exiss beween hem (Engle and Granger, 1987). For F and S o be coinegraed, hey mus be inegraed of he same order. Performing uni roo ess on each price series deermines he order of inegraion. If each series is nonsaionary in he levels, bu he firs differences and he deviaions ε are saionary, hen he prices are coinegraed of order (1,1), denoed CI (1,1), wih he coinegraing coefficien δ 1. In order o es for coinegraion, wo economeric procedures are implemened: he Engle-Granger wo-sep mehodology (Engle and Granger, 1987) and he Johansen s Maximum Likelihood approach (Johansen, 1988 and 1991). According o Engle and Granger, wo basic seps are followed: 1. Tesing he exisence of uni roos (inegraion order) in each index, following Augmened Dickey-Fuller (ADF) es hrough equaion Coinegraion esing beween sock index spo and fuures marke. Consider prices (in log) in spo marke i and fuures marke j ( S and F ), and P is he vecor ha i j 10

11 consiss of S and F. According o Engle and Granger (1987), i j S i is said o be inegraed of order d, denoed i S ~ I (d), if he dh difference of i S is saionary. The vecor P is said o be coinegraed of order d, b, denoed as P ~ CI (d, b), if each componen of P is inegraed of order d, and here exiss a non-zero vecor δ such ha δ P is inegraed of order d-b, for b>0. If boh i S and j F are I (1) and P ~ CI (1,1) [i.e. δ P ~ I (0)], hen here are error-correcion equaions in he following form: i S = α 1 [S i 1 -δ 1 F j 1 ] + lagged ( S i and F j ) + F j = α 2 [F j 1 -δ 2 S i 1 ] + lagged ( S i and F j ) + i e (4) j e where α 1 and α 2 are non-zero coefficiens and e και e i j are saionary, possibly auocorrelaed error erms. Engle and Granger proposed several coinegraion ess; however, he mos preferable is he ADF saisic es. In order o es for coinegraion beween he wo markes, he Johansen s Maximum Likelihood Procedure (Johansen, 1988) is also implemened. This is a preferred mehod of esing for coinegraion as i allows resricions on he coinegraing vecors o be esed direcly, wih he es saisic being x 2 disribued. This specific procedure provides a unified framework of esimaing and esing he coinegraion relaionships in a VAR error correcion mechanism, which incorporae differen shor-run and long-run dynamic relaionships in a variable sysem. The Johansen s procedure firsly specifies he following unresriced N-variable VAR: x + Π µ x ε (5) = k i= 1 i + i 11

12 where x = [ f, s ], µ is a vecor of inerceps erms and ε is a vecor of error erms. Johansen (1988) and Johansen and Juselius (1990) reparameerized eq. (5) in he form: x k 1 i= 1 = µ + Γ x x ε (6) i i + Π k + Equaion (6) is now a VAR reparameerized in error correcion form, where Π= - (Π-Π 1 - -Π k ) represens he long response marix. Wriing his marix as Π = αβ, hen he linear combinaions β x k will be I(0) in he exising of coinegraion, wih α being he adjusmen coefficiens, and he marix Π will be of reduced rank. The Johansen approach can be used o es for coinegraion by assessing he rank (r) of he marix Π. If r=0 hen all he variables are I(1) and here are no coinegraing vecors. If r=n hen all of he variables are I(0) and, given ha any linear combinaions of saionary variables will also be saionary, here are N coinegraing vecors. Las, if 0<r<N here will be r coinegraing vecors. 3.3 Error Correcion Model and Causaliy The coinegraion beween wo series involves a coninuous adjusmen of innovaions prices, so ha hese would no become larger in he long run. Engle and Granger (1987) have shown ha all he coinegraed series can include an error correcion (he Granger represenaion heorem ) and, on he conrary, he exisence of coinegraion is a necessary condiion in order o consruc error correcion models. The accepance ha each pair of cash and fuures prices composes a coinegraing sysem leads o he implemenaion of an error correcion model for each series, which is characerized by he abiliy o overcome problems caused by spurious resuls. 12

13 If S and F denoe he firs differences of he fuures and cash prices, he following coinegraing regressions are possible: S F n n 1 + as z 1 + α11 i) S i + α12( i) F i + i= 1 i= 1 = α ( ε (7) n n 2 + af z 1 + α 21 i) S i + α 22( i) F i + i= 1 i= 1 = α ( ε (8) S F where z = S [b + a F ] is he error correcion erm. Equaions (7) and (8) represen a vecor auoregression (VAR) in firs differences; hus, all variables are held joinly endogenous and OLS is an appropriae mehod of esimaion. Each equaion is inerpreed as having wo pars. The firs par is he equilibrium error (from he coinegraing regressions). This measures how he lef-hand-side variable adjuss o he previous period s deviaion from long run equilibrium. The remaining porions of he equaions are he lagged firs differences, which represen shor-run effecs of he previous period s price changes on he curren period s price changes. For example, in equaion (7) he change in S is due o boh shor-run effecs, possibly from boh which represens adjusmen o long-run equilibrium. Fs and Ss, and o he las-period equilibrium error, z 1, The error correcion erm eners ino he wo equaions wih a one period lag and is esimaed from he coinegraing regressions, wih consan erms being included o make he mean of he error series zero. The coefficiens α S και α F aached o he error correcion erm measures he single-period response of he lef-hand-side variable o deparures from equilibrium ( speed of adjusmen coefficiens ). A leas one speed of adjusmen coefficiens mus be nonzero for he model o be an error correcion model. 13

14 If he value of α S in eq. (7) is zero, he curren period change in he index does no respond a all o he las period s deviaion from long-run equilibrium. The link beween coinegraion and causaliy sems from he fac ha if spo and fuures indices are coinegraed, hen causaliy mus exis in a leas one direcion (unidirecional causaliy) and possibly, in boh direcions (bi-direcional causaliy) (Granger, 1986). Since coinegraion implies ha each series can be represened by an error correcion model ha includes las period s equilibrium error, as well as lagged values of he firs differences of each variable, emporal causaliy can be assessed by examining he saisical significance and he relaive magniudes of he error correcion coefficiens and he coefficiens on he lagged variables (Wahab and Lashgari, 1993). The Sandard Granger causaliy ess do no ake ino accoun he significance of error correcion coefficiens. Engle and Granger (1987) focused on he fac ha he esimaes of a VAR are misspecified in he case of coinegraed variables, because he error correcion erms ha are aached o error correcion models are no accouned. The argumen is ha he models implemened for esing he causaliy relaionship are misspecified if he variables, which are esed for he direcion of causaliy, are coinegraed. The exisence of a unidirecional causaliy from S o F requires: (i) ha some of he a 21 s in eq. (8) mus be non-zero while all he a 12 s in eq. (4) mus be equal o zero and/or (ii) he error correcion coefficien α F in eq. (8) is saisically significan a convenional levels. If he coefficiens a 21 and a 12 are individually and joinly non-zero, hen a feedback relaionship or a bi-direcional causaliy beween he wo price series is exised. On he oher hand, if he above coefficiens are equal o zero, hen here is no a 14

15 causaliy relaionship beween he wo variables, as each variable is deermined by is prices and he relevan innovaions. 4. DATA AND EMPIRICAL RESULTS Price daa on he FTSE/ASE-20 sock index and he hree-monh FTSE/ASE-20 index fuures conrac are from he Ahens Sock Exchange (ASE) and he Ahens Derivaives Exchange (ADEX) respecively. Daily daa are used during he period from Augus 1999 unil June The logs of he spo and fuures prices are used. The fuures prices are always hose of he nearby conrac. To avoid hin markes and expiraion effecs, we rollover o he nex nearby conrac one week before he nearby expires. Moreover, in order o eliminae he sale price effecs, prices before and afer he specified rading hours of ASE and ADEX are no used. Afer he esablishmen of ADEX and from Augus 1999 unil March 2000, average monhly rading volume in FTSE/ASE-20 fuures conracs has risen 76.6% ( conracs in March 2000), while daily average number of conracs in March 2000 has increased 131% relaive o Augus 1999 (2.816 and conracs in March 2000 and Augus 1999 respecively). The selecion of he esimaion period s lengh is due o he significan increase of a number of saisics concerning he FTSE/ASE-20 fuures conrac in 2001 compared o he previous year. Table 2 repors oal volume, daily average volume and daily open ineres on FTSE/ASE-20 index fuures during he period Average daily raded volume for 2001 was up o 173% compared o 2000, open ineres averaged over conracs, while daily average rading value 15

16 for 2001 was 41,76 mil. euros. These saisics indicae he srong growh of so far he sar produc of ADEX 2. TABLE 2: Main Indicaors: FTSE/ASE-20 Index Fuures Year % change from 2000 Trading days Toal volume 1.320, Daily average volume % Daily average open ineres % Traded value (in mln euros) % Source: ADEX To deermine he order of each price series, he Augmened Dickey-Fuller τ-es and he Phillips-Perron es are compued on he levels of each price series. Performing he ess on he levels of each series shows ha he null hypohesis of a uni roo is no rejeced; hus, each series is I(0). On he conrary, he resuls of he ess οn he firs differences indicae ha each series is I(1). Table 3 repors he resuls of he Uni roos ess. 2 According o Federaion of European Securiies Exchanges, ADEX ranked 7h in sock index fuures by rading value during among European derivaives markes, leaving behind he markes in Porugal, Denmark, Finland, Ausria, Norway, Poland, and Hungary. 16

17 TABLE 3: Uni Roo Tess Saisic ess Spo index Fuures index ADF levels ADF firs differences * * PP levels PP firs differences * * The null hypohesis is ha series has a uni roo. *Denoes ha he es saisics are significanly differen from zero a he 5% level. The criical value for ADF and PP ess is a he 5% level. Since he wo series are I(1), boh he Engle-Granger s ess and he Johansen s procedure ess for coinegraion are used. Engle-Granger s coinegraion ess are implemened o he residuals of he bivariae regressions. Table 4 repors he resuls of DF and ADF ess 3. The resuls indicae he exisence of a saisically significan coinegraion relaionship beween he wo markes. 3 Analysis of he daa indicaed he presence of non-normaliy. The problem of non-normaliy in he daa is overcome by including a dummy variable relaing o a specific observaion in each index. The resuls in Table 4 and 5 relae o ess including saionary dummy variables. Exclusion of he dummies does no aler he paern of resuls. 17

18 TABLE 4: Engle- Granger Coinegraion Tess Sysem DF ADF Spo ; Fuures * (5) Fuures ; Spo * (5) On each sysem, he firs marke is he dependen variable (on he lef of he sign ;), while he oher marke is he independen variable (on he righ of he sign ;). * Denoes ha he null hypohesis of no coinegraion is rejeced a he 5% level. The criical value is a he 5% level. The number in parenheses shows he leas required lag order o have whie noise innovaions. Deparing from he bivariae coinegraion regressions in he Engle-Granger framework, a VAR error coinegraion model such as in Equaion (6) is esimaed in order o consider he wo series joinly and cross check he exisence of coinegraion beween hem, according o he procedure advanced by Johansen. Hall (1991) has demonsraed ha in using his procedure o es for coinegraion i is necessary o carry ou ess o esablish he appropriae order of he VAR. The choice of opimal lags is given by consideraion of minimizing he Akaike informaion crierion (Akaike, 1973) and absence of auocorrelaion in he VAR residuals; five lags for he levels of he variables are included. Table 5 repors he Likelihood raio (LR) es for coinegraion based on Maximal eigenvalue and Trace es saisics for he number of coinegraing vecors. The es saisics for he alernaive hypohesis r 1 are greaer han he criical values a he 5% level. These resuls indicae ha he null hypohesis of zero coinegraing vecors is rejeced a 5% level, whils he hypohesis of one coinegraing vecor canno 18

19 be rejeced. Thus, he spo price level and he fuures price level are I(1), wih linear combinaions being I(0), confirming ha he wo price series are CI(1,1) 4. TABLE 5: Johansen Tess for Coinegraion of Spo and Fuures Prices LR Tes for coinegraing vecors based on Variables Null Hypohesis Alernaive Hypohesis Maximal eigenvalue Trace Eigenvalues LS, LF r=0 r 1 r 1 r= * * LS and LF denoe he log of spo and fuures prices respecively. * Denoes ha he null hypohesis of no coinegraion is rejeced, while he alernaive hypohesis of one coinegraing vecor is acceped a he 5% level. The criical value for Maximal eigenvalue and Trace saisics is and a he 5% level respecively. Since boh price series are CI(1,1), an error correcion model (ECM) wih lag lengh 5 on S and F is esimaed using OLS regression 5 : 4 The robusness of he resuls using he Johansen procedure in relaion o violaions of non-normaliy and heeroskedasiciy was examined using error-based es for coinegraion proposed by Phillips and Perron (1988). The resuls here were confirmed. 5 The full oupu from he esimaion of he error correcion model is available from he auhor upon reques. 19

20 LS LF = a = a as z ( i) LS i + α12( i) LF i + i= 1 i= 1 α ε (9) af z ( i) LS i + α 22( i) i= 1 i= 1 α LF ε (10) i + S F In esimaing he ECM, we faced wih he problem of serial correlaion and heeroskedasiciy. So, he OLS esimaion is carried ou using Cochrane-Orcu wo sep mehod auo-regressive processes (AR) and Whie (1980) correcion for heeroskedasiciy o accoun for hose problems. Table 6 displays he es resuls of he resricions imposed on he speed of adjusmen coefficiens (α S and α F ) and he lagged variables coefficiens (α 12 and α 21 ) o eq. (9) and (10), using he Wald es saisic, which being x 2 disribued. TABLE 6: Wald Tes Resuls Null Hypohesis (H 0 ) Wald saisic P value α S = α F = α 12 = α 21 = α 12 = 0, α 21 = The resuls of he Wald es on he speed of adjusmen coefficiens (α S and α F ) indicae ha he spo and fuures conrac behave somewha differenly. The lack of significance of α S means he spo marke does no respond o he previous period s deviaion from equilibrium. The significance of α F means he curren period fuures innovaion responds o he previous period s deviaion from equilibrium. The finding 20

21 ha one of he speeds of adjusmen coefficiens is nonzero (α F 0) confirms ha he model is an error correcion model. The significan speed of adjusmen in eq. (10) does no mean ha he spo marke leads or causes he fuures marke. Respecively, he insignifican speed of adjusmen in eq. (9) does no mean ha he fuures marke is no leading he spo marke. In order o conclude abou he direcion of causaliy or he lead-lag relaionship beween he wo markes, we have o es he significance of he lagged variables coefficiens. The resuls of he Wald es on coefficiens α 12 and α 21 show ha he null hypohesis (he coefficiens are individually and joinly equal o zero) canno be acceped. Thus, he significance of α 12 and α 21 indicaes he exisence of a bi-direcional causaliy or a feedback relaionship beween he wo markes, since he las period s price changes in S (F ) shor run affec he curren period s price changes in F (S ). This finding indicaes ha he FTSE/ASE- 20 fuures conrac serves as a focal poin of informaion assimilaion and fulfills is price discovery funcion. So, fuures prices conain useful informaion abou subsequen spo prices, beyond ha already embedded in he curren spo price. This has an imporan implicaion for marke paricipans in he Greek capial marke, indicaing ha here are opporuniies for significan arbirage profis and hedging sraegies. Finally, our finding ha a newly esablished (emerging) derivaives marke provides he funcion of price discovery deserves furher discussion. There are many reasons which migh explain why fuures prices lead cash index prices. The firs explanaion is ha he fuures marke is less cosly for raders o uilize han he cash marke. Oher reasons could be he lower ransacions coss in he fuures marke, he 21

22 ease in shoring fuures conracs and he invesors preference o hold fuures conracs because hey are no ineresed in he underlying asse per se. 5. SUMMARY AND CONCLUSIONS This paper examines he relaionship beween he FTSE/ASE- 20 sock index and is hree-monh fuures conrac during he period from Augus 1999 unil June The aim is o invesigae he price discovery in he Greek FTSE-20 index fuures marke and o deermine he informaional linkage beween he spo and he fuures marke. The uni roo ess conclude ha each series is non-saionary in he levels bu saionary afer firs differencing. Boh he spo index and he fuures markes are esed for coinegraion using boh he Engle-Granger and Johansen mehods. Boh esing procedures indicae ha he wo markes are coinegraed. Thus, an error correcion model is developed in order o invesigae he causaliy or he lead-lag relaionship beween he wo markes. The resuls of his model indicae he presence of a bi-direcional causaliy beween he spo index and he fuures index markes, and hus an informaional linkage beween hem. Tha means he index prices in fuures marke (cash marke) may conain useful informaion regarding consequen price movemens of he sock index marke (fuures marke). This empirical finding suggess ha he newly esablished ADEX marke provide fuures conracs ha can be used as vehicles of price discovery and indicaes he imporan role ha his fuures marke plays in he Greek capial marke owards is ulimae mauriy, ransparency and secure funcioning. The exisence of such an informaional linkage beween sock index spo and fuures markes implies ha 22

23 invesors using hese markes can explore significan arbirage profis and hedging opporuniies. Finally, his paper provides wo direcions for fuure research regarding he price discovery in he ADEX. Firs, he relaionship beween price discovery and volume of fuures rading in each monh of he FTSE/ASE-20 conrac consiues an ineresing opic for research. Second, aenion has o be given o he relaionship in volailiies beween he wo markes. If volailiy spillovers exis from one marke o he oher, hen he volailiy ransmiing marke may be used as a vehicle of price discovery, since such informaion may conribue o he decision-making process. REFERENCES Akaike, H. (1973), Informaion Theory and he Exension of he Maximum Likelihood Principle, Ιn B. N. Perov and F. Caski, eds. (Second inernaional symposium on informaion heory, Budapes). Anoniou, A. and I. Garre (1993), To Wha Exend did Sock Index Fuures Conribue o he Ocober 1987 Sock Marke Crash?, The Economic Journal 103, Booh, G.G., R.W. So, Y. Tse (1999), Price Discovery in he German Equiy Index Derivaives Markes, Journal of Fuures Markes 19, Chan, K. (1992), A Furher Analysis of he Lead-lag Relaionship beween he Cash Marke and Sock Index Fuures Markes, Review of Financial Sudies 5,

24 Cochrane, D. and G.H. Orcu (1949), Applicaion of Leas Squares Regression o Relaionships Conaining Auocorrelaed Error Terms, Journal of he American Saisical Associaion 44, Dickey, D.A. and W.A. Fuller (1981), Likelihood Raio Saisics for Auoregressive Time Series wih a Uni Roo, Economerica 49, Elam, E. and D.B. Dixon (1988), Examining he Validiy of a Tes of Fuures Marke Efficiency, Journal of Fuures Markes 8, Engle, R.F. and C.W.J. Granger (1987), Coinegraion and Error Correcion Represenaion, Esimaion and Tesing, Economerica 55, Granger, C.W.J. (1986), Developmens in he Sudy of Coinegraed Economic Variables, Oxford Bullein of Economics and Saisics 48, Grübichler, A., F.A. Longsaff, E. Schwarz (1994), Elecronic Screen Trading and he Transmission of Informaion: An Empirical Examinaion, Journal of Financial Inermediaion 3, Hall. S.G. (1991), The Effec of Varying Lengh VAR Models on he Maximum Likelihood Esimaes of Coinegraing Vecors, Scoish Journal of Poliical Economy 38, Iihara, Y., K. Kao, T. Tokunaga (1996), Inraday Reurn Dynamics beween he Cash and Fuures Markes in Japan, Journal of Fuures Markes 16 (4), Johansen, S. (1988), Saisical Analysis of Coinegraion Vecors, Journal of Economic Dynamics and Conrol 12, Johansen, S. (1991), Esimaion and Hypohesis Tesing of Coinegraion Vecors in Gaussian Vecor Auoregressive Models, Economerica 59 (6),

25 Johansen, S. and K. Juselious (1990), Maximum Likelihood Esimaion and Inference on Coinegraion wih Applicaions o he Demand for Money, Oxford Bullein of Economics and Saisics 52, Kawaller, K.G., P.D. Koch, T.W. Koch (1987), The Temporal Price Relaionship beween S&P 500 Fuures and he S&P 500 Index, Journal of Finance 42 (Dec.), Ng, N. (1987), Deecing Spo Prices Forecass in Fuures Prices using Causaliy Tess, The Review of Fuures Markes 6, Pizzi, M., A. Economopoulos, H. O Neil (1998), An Examinaion of he Relaionship beween Sock Index Cash and Fuures Markes: A Coinegraion Approach, Journal of Fuures Markes 18 (3), Soll, H.R. and R.E. Whalley (1990), The Dynamics of Sock Index and Sock Index Fuures Reurns, Journal of Financial and Quaniaive Analysis 25, Shyy, G., V. Vijayraghavan, B. Sco-Quinn (1996), A Furher Invesigaion of he Lead-lag Relaionship beween he Cash Marke and Sock Index Fuures Marke wih he use of Bid/Ask Quoes: The Case of France, Journal of Fuures Markes 16 (4), Phillips, P. and P. Perron (1988), Tesing for a Uni Roo in Time Series Regression, Biomerica 75, Tang, Y.N. and Y.K. Ho (1989), The Pricing of Hang Seng Index Fuures and Causaliy beween Fuures and Spo Prices, Paper presened a he Annual Meeing of he Academy of Inernaional Business, Nov , Singapore. 25

26 Wahab, M. and M. Lashgari (1993), Price Dynamics and Error Correcion in Sock Index and Sock Index Fuures Markes: A Coinegraion Approach, Journal of Fuures Markes 13 (7), Whie, H. (1980), A Heeroskedasic-Consisen Covariance Marix Esimaor and a Direc Tes for Heeroskedasiciy, Economerica, May: