Index futures trading, information and stock market ... · futures trading activity on the volatility of the FTSE 100 index. Antoniou et al.,23 however, examine further the impact

Abstract

This paper examines the effect of futures tradingon the volatility of the underlying spot market.It focuses on various techniques to investigate therelationship between information and thevolatility of the FTSE/ASE-20 and

FTSE/ASE Mid 40 indices in Greece. Theresults for the FTSE/ASE-20 index suggestthat futures trading has led to decreased stockmarket volatility (negative effect), but the resultsfor the FTSE/ASE Mid 40 index indicate thatthe introduction of stock index futures has led to

146 Derivatives Use, Trading & Regulation Volume Twelve Numbers One/Two 2006

Index futures trading, information and stock

market volatility: The case of Greece

Christos Floros* and Dimitrios V. Vougas**

*Department of Economics, University of Portsmouth, Portsmouth Business School,Portsmouth, PO1 3DE, UK. Tel: +44 (0)2392 844244;E-mail: [email protected]**Department of Economics, University of Wales Swansea, Singleton Park, SwanseaSA2 8PP, UKTel: +44 (0)1792 602102; E-mail: [email protected]: (in revised form) 4th May, 2006

Christos Floros completed his first degree in mathematics and operational research at Brighton University andalso holds MA (economics) and MSc (mathematics) degrees from Portsmouth University and a PhD in financialeconomics from Swansea University. He is a lecturer in banking and finance at the University of Portsmouthand his research interests include financial econometrics, derivatives and banking efficiency.

Dimitrios V. Vougas studied economics and econometrics at the Athens University of Economics andBusiness and the University of Bristol with a scholarship from the State Scholarship Foundation of Greece. Helectures at Swansea University and previously held a research position at the University of Nottingham andlecturing positions at Middlesex and Portsmouth universities.

Practical applications

The introduction of a futures market and, in particular, the impact of futures on stockmarket volatility is a long debate. Previous studies show that the futures market leads to anincrease in market depth and a decrease in volatility. This is due to the more rapid rate atwhich information is reflected in prices and speculation. Other studies suggest that adecrease in cash market volatility is due to an increase in market liquidity. Empirical studiesfor UK and US financial markets do not conclude clearly whether the introduction offutures stabilises or destabilises the underlying spot market. It is therefore important forpractitioners to look at the link between information (news) and volatility. To the authors’knowledge, this is the first study that examines this effect of the Greek futures market onstock market volatility.

Derivatives Use,

Trading & Regulation,

Vol. 12 No. 1/2, 2006,

pp. 146–166

� Palgrave MacmillanLtd1747–4426/06 $30.00

increases through speculation or arbitragestrategies. Then, an increase in interest ratesand the cost of capital, leading to areduction in the value of investments, isquite possible. From a financial viewpoint,the price volatility usually depends on thearrival of new information in the market. Ifthe market is efficient, the price reflects thisnew information, but the literature presentsarguments that futures markets increasemarket depth2 and reduce spot marketvolatility. One possible explanation whyfutures markets reduce the spot marketvolatility is financial risk.

The majority of papers in the literatureconsider the effect of futures trading onstock market volatility using either anunconditional measure of volatility orGARCH models. GARCH models spotand futures market volatility. This paperuses both techniques. In addition, most ofpapers in the literature look at data fromhighly developed markets such as the USand UK.

This paper examines two newlyintroduced Greek stock index futures onthe FTSE/ASE-20 and FTSE/ASE Mid 40.It addresses any stabilisation/destabilisationeffects and the impact of futures trading onthe associated stock market indices. In thisrespect, the findings are very important inthe absence of any previous workexamining the effect of futures trading onthe Greek spot index markets.

The plan of the paper is as follows. Inthe second section, the theoretical debateand literature review are presented. Thethird section outlines the methodology,while the fourth section describes the data.The fifth section presents empirical results

increased volatility (positive effect), while theestimations of the unconditional variancesindicate lower market volatility after theintroduction of stock index futures. Furthermore,the results show that good news has a morerapidly impact on FTSE/ASE-20 stock returnvolatility. For the FTSE/ASE Mid 40 index,the results suggest that news is reflected in pricesmore slowly, while old news has a less persistenteffect on prices. These findings are helpful tofinancial managers dealing with Greek stockindex futures.

INTRODUCTION

The debate on the impact of futures onstock market volatility is still controversial.In other words, the issue of whether thefutures markets affect underlying spotmarkets is not widely accepted. Some pastcritics of index futures agree that theintroduction of stock index futures increasesstock market volatility. Others report nosignificant volatility effect associated withthe introduction of stock index futures. Ingeneral, the link between futures tradingand stock market volatility has become verycomplex. Several studies have empiricallyexamined the long-term relationship acrossthe decades. Most of them compare thevolatility of the spot market before andafter the introduction of futures trading,using econometric models. According toBologna and Cavallo,1 there are twotheories in the literature about therelationship between futures markets andunderlying spot markets. The first theorysupports the argument that futures tradingdestabilises the underlying spot market byincreasing its volatility. In practice, volatility

147Floros and Vougas

from various econometric models, and thefinal section concludes the paper andsummarises the findings.

THEORETICAL DEBATE AND

LITERATURE REVIEW

Futures contracts are attractive to investorsand speculators because they have lowmargin requirements and low transactioncosts. The above does not imply, however,that futures are destabilising. Apparently,several theoretical arguments have beenused over the years to explain whyderivatives markets in general might affectthe volatility of the spot market (theunderlying asset market). Theoretically,when futures are introduced, one can buyor sell futures, options, commodities orother products. This may influence spotprices, if the producers respond to newinformation. Various information-basedmodels analyse and explain whether theopening of a futures market may stabilise ordestabilise stock prices. This literatureincludes papers by Stein,3 Subrahmanyam4

and Newbery.5

In addition, according to Chari et al.6

futures markets may stabilise or destabiliseprices. Since futures and spot prices areusually closely related, futures are attractiveto speculators. Antoniou and Holmes7 arguethat the arrival of futures trading dependson speculators’ information. In other words,when speculators have perfect information,the introduction of futures markets stabilisesprices. But when speculators have a noisysignal, there is a destabilising effect. Insome cases, however, the introduction of afutures contract leads to a more stable spotprice.

Since futures markets have a higherdegree of leverage than cash markets do,they are more likely to attract uninformedspeculative investors and thus destabilisecash markets by increasing volatility. Butfutures markets increase the overall marketdepth and informativeness, are importantfor price discovery, allow the transfer of riskand may actually reduce spot volatility.

Index futures contracts have severaladvantages over the underlying stock index.These enable holders to trade on accessiblecredit and reduce costs using severalstrategies. Index futures have somedisadvantages, however, such as anon-beneficial increase in price variability;see Hodgson and Nicholls.8

Several studies examine the relationshipand behaviour between stock marketvolatility and index futures volatility. Forinstance, Edwards9 examines stock marketvolatility before and after the introductionof futures. He finds that the introduction ofthe S&P500 futures contract has reducedcash market volatility. Hodgson andNicholls,8 however, examine the effect ofindex futures on Australian sharemarketvolatility, and argue that the introduction oftrading in index futures has not affected thelong-term volatility of the underlying spotshare market. Further, Chang et al.10 pointout that futures trading increases spotportfolio volatility (with no volatilityspillovers). In addition, Aggarwal11 andHarris12 find that the post-futures period ismore volatile. Harris12 reports that thevolatility of S&P500 stocks increased,relative to the volatility of stocks in acontrol system. Also, Lockwood and Lim13

find that cash stock market volatility

148 Floros and Vougas

argue that it is better to model thevolatility following a GARCH (1,1) processas the conditional variance of the errorterm is a linear function. Hence, they use aGARCH (1,1) process with a dummyvariable to investigate the impact of futurestrading on spot market volatility. Theirresults show that the volatility of the spotmarket with futures is greater than thatwithout futures. For the whole period, theanalysis with GARCH (1,1) suggests thatthere is an increase in spot price volatilityon futures. For the sub-periods, theGARCH parameters are significantlydifferent from zero and, also, for both pre-and post-futures trading there is an increasein the unconditional variance, and thereforean increase in volatility and a definiteimpact of trading in futures on spot pricevolatility has been found. Furthermore,Brorsen19 and Gulen and Mayhew20 suggestthat volatility increases in many markets.Gulen and Mayhew20 examine stock marketvolatility before and after the introductionof equity index futures trading in 25countries using the GJR-GARCH, theNon-linear GARCH and the EGARCHmodels. In addition, Holmes21 examines theimpact of FTSE Eurotrack futures tradingon spot volatility using daily closing priceindices for the period June 1990–April1994. His results show that futures contractshave a beneficial impact on price discoveryin the spot market, even when tradedmarket is thin. By contrast, Robinson22

provides evidence of a stabilising effect offutures trading activity on the volatility ofthe FTSE 100 index. Antoniou et al.,23

however, examine further the impact offutures trading on FTSE 100 stock index

increased after stock index futures trading.Also, Maberly et al.14 find that volatilityincreased in the S&P500 index.

Further, for the New York StockExchange (NYSE), Chang et al.15 examinethe effects on futures volatility and pricechanges. Empirical results show that there isan increase in volatility only at the close ofthe futures market. By contrast, when theNYSE closes, volatility falls significantly.Another interesting point is that highervolatility in the spot market increases thedemand for further hedging in stock indexfutures.16

Since analysis of the introduction ofindex futures usually depends on the choiceof pre- and post-futures periods,14 the effectof future trading on volatility can bemodelled using GARCH-family models.Specifically, one can examine whether theexistence of futures trading has any effecton volatility, using a conditional variancemodel with a dummy variable.

First, Chan et al.17 examine the intra-dayvolatility spillovers between S&P500 indexstock and futures markets using GARCHmodels and show a strong cross-marketdependence in the volatility process. Then,Antoniou and Foster18 analyse the effect offutures trading on Brent crude oil spotprice volatility (for 1988). From GARCHanalysis, they find an increase ininformational spot market efficiency and adecrease in the old-news term.

Another approach is by Antoniou andHolmes.7 They use GARCH models toexamine the relationship between trading,information and volatility in the FTSE 100stock index futures (using data fromNovember 1980 to October 1991). They


volatility. Now using the GJR-GARCH(1,1) model with a dummy variable, theyconclude that futures trading has a negativeimpact on the volatility of the stockmarket. Further, Butterworth24 examinesthe link between volatility and news for theFTSE Mid 250 index using symmetric andasymmetric GARCH methods. He arguesthat the new information after the onset offutures is reflected in prices less rapidly and,therefore, an increase in the persistence ofvolatility is quite possible.

In addition, Rahman25 examines theimpact of index futures trading on thevolatility of component stocks for the DowJones Industrial Average (DJIA). The datacover the period April–June 1997(pre-futures) and April–June 1998(post-futures). Rahman25 uses the simpleGARCH (1,1) model to estimate theconditional volatility of intra-day returns.The empirical results confirm that there isno change in conditional volatility frompre- and post-futures periods.

Another good representation for volatilitymodelling is the EGARCH model.Koutmos and Tucker26 use the EGARCHmodel and show that, for both stock andfutures markets, bad news increasesvolatility more than good news. Further,Pericli and Koutmos27 use an EGARCHmodel and find that the volatility of theS&P500 index has decreased after theintroduction of futures trading (forS&P500). Darrat et al.28 examine thevolatility and test the relationships betweenvolatilities in the spot and futures markets.The sample spans the period after the stockmarket crash of October 1987 (ieNovember 1987–November 1997). They

use an EGARCH model to measure spotreturns volatility and futures returnsvolatility. According to their empiricalresults, ‘futures trading is not responsible forcash market volatility’. Therefore, a possibleexplanation is that an unstable spot marketcould affect volatility in the index futuresmarket.

Further, Lien and Tse29 examine theeffect and the relationship between cashsettlement and cash-futures prices using abivariate GARCH model with a dummyvariable (for measuring those effects). Theirempirical results show that the futuresmarket leads the cash market and that thecash settlement has never had an impact oncash and futures returns (ie the coefficientof the dummy variable is found to bestatistically insignificant). In addition,Kyriacou and Sarno30 examine therelationship between UK spot marketvolatility and futures trading using theGARCH process. The time series spansfrom 1 October 1992 to 29 December1995. The empirical results suggest thatspot market volatility could be modelled bya GARCH (1,1), and volatility ischaracterised as statistically reliable.Kyriacou and Sarno30 argue that futurestrading affects spot market volatility in‘opposite directions’.

Gulen and Mayhew20 examine the timeseries behaviour of stock indices for 25countries, using several GARCH models.As they point out, in the US and Japan,futures trading leads to an increase inconditional volatility. Bologna and Cavallo1

examine the effect of the introduction ofstock index futures, that is whether itreduces stock market volatility (for Italian


the trading of MSCI Taiwan futures has noeffects except the asymmetric responsebehaviour. Bae et al.34 examine the effect ofthe introduction of index futures trading inthe Korean markets on spot price volatility.The results show that there is an increasein spot price volatility after the introductionof futures trading. These findings are in linewith Ryoo and Smith35 for the Koreanmarket. They find that futures tradingincreases the speed at which information isreflected in spot market prices, reduces thepersistence of information and increases spotmarket volatility. Recently, Antoniou et al.36

test the hypothesis that the introduction ofindex futures has increased positivefeedback trading in the spot markets of sixindustrialised nations. Their findings supportthe view that futures markets help stabilisethe underlying spot markets by reducingthe impact of feedback traders andattracting more rational investors.

Overall, it is evident that the majority ofstudies discussed conclude that futures mayhave either a positive or a negative impacton stock market volatility. Also, a decrease orincrease in volatility after the introduction offutures is found. Hence, the debate on theimpact of futures trading on stock marketvolatility does not give clear conclusions.

To analyse the impact of futures tradingon Greek stock market volatility, severalGARCH models are estimated for bothsub-periods, thereby allowing a comparisonof volatility before and after theintroduction of futures trading. SeveralGARCH models are used to find whichmodel fits the data best. Also, one needs toknow how these GARCH specificationsmodel volatility before and after the

market). Their empirical results show thatthe introduction of stock index futuresaffects the volatility of the spot market.Also, the results from GARCH (1,1)models — for pre-futures and post-futuressub-periods — suggest that the indexfutures market reduces volatility. Inaddition, the ‘old news’ parameter from thepresence of stock index futures trading hashad a different (less) impact for stockmarket volatility estimation.

Pilar and Rafael31 analyse the effect ofthe introduction of futures in the Spanishstock market using a GJR model. Theresults indicate a decrease in volatility afterthe introduction of futures. Anotherempirical approach on the effect of futurestrading on volatility is that fromFiguerola-Ferretti and Gilbert.32 They useAluminium transaction prices to examinewhether futures trading affects volatility. Todo so, they use error-correction models andthe GARCH(1,1) regression model. Inaddition, they report the results of a VARmodel and display impulse response analysisto track the effects of a shock to each ofthe volatilities. Their results show thatvolatility decreases in the post-period.

Chiang and Wang33 examine the impactof futures trading on Taiwan spot indexvolatility. This study also discusses themacroeconomic effects on spot volatilityand the asymmetric effects of futurestrading on spot price volatility behaviour.The data cover the period from 5 January1995 to 10 March 2000. The asymmetrictime-varying GJR volatility model isemployed. Their empirical results show thatfutures trading of Taiwan Index Futures hasmajor impacts on spot price volatility, while


introduction of futures trading. Notice thatGARCH models link information (news) tovolatility.

METHODOLOGY

Past empirical papers show that the effect offuture trading on volatility can be modelledusing conditional variance models. In otherwords, one can examine whether theexistence of futures trading has any effecton volatility by using a (GARCH) volatilitymodel with a dummy variable (taking thevalue zero pre-futures and onepost-futures). If the dummy is statisticallysignificant, the existence of futures tradinghas an impact on spot market volatility (seeAntoniou et al.23).

Many studies analyse the ‘futures effect’using the standard GARCH(1,1) model1,24

or the GJR model, which tests for thepresence of asymmetries.23,24,31

To analyse the effect of future trading onstock market volatility, GARCH models arealso employed. The GARCH(p,q) modelcaptures better the tendency of returns toexhibit volatility clustering, incorporatingheteroscedasticity into the estimationprocedure. In this model, positive andnegative past values have a symmetric effecton the conditional variance. The standardGARCH(1,1) model with dummy variablecan be expressed as follows

Rt � � � �t

� 2t � � � a1�

2t–1 � �1�

2t–1 � cDi � u (1)

The dummy variable Di takes zero valuefor the pre-futures period and one for thepost-futures period. This dummy allows

one to determine whether futures prices arerelated to any change in the spot marketvolatility. When the coefficient of thedummy variable is positive (negative), thereis a positive (negative) effect of futurestrading on volatility. In addition, assumingthat markets are efficient, a1 (the ARCHparameter) can be viewed as a ‘news’coefficient, while �1 (the GARCHparameter) can be viewed as ‘old news’ andthe persistence coefficient (see Antoniouand Holmes7). Furthermore, according toButterworth24, an increase (decrease) in a1

suggests that news is reflected in pricesmore rapidly (slow). A reduction in �1

suggests that old news has a less persistenteffect on prices changes. In addition, anincrease in �1 suggests greater persistence.Also, when the sum a1 � �1 approachesunity, volatility shocks are persistent.

Other specifications of the GARCH(p,q)include the exponential GARCH(EGARCH) and threshold GARCH(TGARCH). Both models capture volatilityasymmetry.

• EGARCH model. The variance equationof Exponential GARCH(1,1) model witha dummy variable is the following

log(� 2t ) � � � a1� �t–1

�t–1�� a2

�t-1

�t–1

� a3 log(� 2t–1) � cDi (2)

• TGARCH(1,1) model. The varianceequation with a dummy variable is givenby

� 2t � � � a1�

2t–1 � a2�

2t–1dt–1

� a3�2t–1 � cDi (3)


Also when a2 > 0, the leverage effect exists.When a2 0 and significant, the newsimpact is asymmetric.

The methodology is based on thefollowing two parts: First, we test for theimpact of futures trading on stock marketvolatility using the above models with adummy variable. The next step consists ofexamining and comparing the volatilityparameters across two sub-periods: thepre-futures period and the post-futuresperiod. GARCH models are estimated withthe Marquardt algorithm, using theHeteroscedasticity Consistent Covarianceoption (see also Floros and Vougas37). Theconditional mean of the data is alsoestimated using ARMA(p,q) models (whennecessary). Both orders are determined bythe AIC. Finally, the unconditional varianceis also estimated.

DATA DESCRIPTION

Daily closing prices for the FTSE/ASE-20index are used over the period September1997–July 2001. For the FTSE/ASE Mid40 index, the daily closing prices are usedover the period December 1999–July 2001.Closing prices for stock indices wereobtained from Datastream.

Next, the main statistics (mean andstandard deviations) of the returns for thesub-periods before and after theintroduction of futures trading are reported.Table 1 contains information for theFTSE/ASE-20 index, and Table 2 for theFTSE/ASE Mid 40 index.

It is clear that daily standard deviations(SD) have changed little. For both periodsbefore and after the introduction of futures

Good news (�t < 0) and bad news (�t > 0)have an impact on a1 and a1 � a2,respectively. In other words, a negativeinnovation (shock) has a greater impactthan a positive innovation on volatility.


Table 1: Statistics for FTSE/ASE-20 logreturns

Sample period N Mean SD

Pre-futures 502 0.002084 0.024643

Post-futures 500 –0.001346 0.019884

Table 3: Results of GARCH models forFTSE/ASE-20 index

Coeff. on Model dummy t-ratio

MA(1)-TGARCH –5.48E-05 –2.0743*

MA(1)-GARCH (1,1) –5.34E-05 –2.1660*

MA(1)-EGARCH –0.1070 –1.9461*

GARCH (1,1) –5.71E-05 –2.2703*

EGARCH –0.1134 –2.1201*

TGARCH –6.06E-05 –2.2261*

* Significant at the 5 per cent level.

Table 2: Statistics for FTSE/ASE Mid 40 logreturns

Sample period N Mean SD

Pre-futures 36 –0.003228 0.037806

Post-futures 393 –0.002826 0.023835

trading the SD falls slightly. This means thatfutures trading may not destabilise theunderlying spot market. A more detailedempirical investigation needs to be carriedout, however, using GARCH-volatilitymodels.

EMPIRICAL RESULTS

FTSE/ASE-20 index

In Table 3, all GARCH models with adummy variable are reported. There is anegative effect of futures trading on stockmarket volatility. The negative effect isstatistically significant and, therefore, thereis a decrease in volatility associated with theintroduction of futures.

The results presented in Table 3 showthat the introduction of FTSE/ASE-20stock index futures has an effect on thevolatility of the underlying spot market.This result is in line with the finding of

Bologna and Cavallo1 for the Italian StockExchange and Antoniou et al.23 for severalstock indices from different countries.

The next step is to examine and comparethe values of volatility parameters for thepre-futures and the post-futures periods.

The results from the MA(1)-TGARCHmodel are presented in Table 4. It is clearthat some GARCH parameters arestatistically significant at the 5 per cent level(with the exception of the ARCHparameter). In the pre-futures period, thenews parameter is not statisticallysignificant, while the leverage effect exists.Therefore, negative shocks have a greaterimpact on volatility. In the post-futuresperiod, however, the leverage effect term isnot significant, indicating that there is noasymmetric effect. The leverage effectchanges from 0.3387 (pre-futures) to 0.1479(post-futures). According to Table 4, therehas been an increase in both ARCH andGARCH parameters. An increase in the


Table 4: FTSE/ASE-20 index

A. Pre-futures period: MA(1)-TGARCH B. Post-futures period: MA(1)-TGARCHMean equation Coefficient t-statistic Mean equation Coefficient t-statistic

� 0.0017 1.5196 � –0.0018 –2.0068*

MA(1) 0.1893 4.1803* MA(1) 0.1616 3.3140*

Variance equation Variance equation

� 0.0001 2.8118* � 7.16E-05 2.9744*

a1 0.0228 0.4726 a1 0.1364 1.7603*

a2 0.3387 2.9968* a2 0.1486 0.9729

a3 0.4990 3.7119* a3 0.6114 6.4064*

*Significant at the 5 per cent level.


Tabl

e 5:

Est

imat

ion

resu

lts o

f G

AR

CH

mod

els

for

FTS

E/A

SE-

20 in

dex

A.P

re-fu

ture

s pe

riod

Mod

els

�a 1

a 2a 3

�1

MA

(1)-

GA

RC

H(1

,1)

6.00

E-0

5 (2

.466

1)*

0.10

21 (

2.52

13)*

0.79

94 (

12.8

99)*

MA

(1)-

EG

AR

CH

–2.7

736

(-2.

6371

)*0.

2530

(2.

7142

)*–0

.199

9 (-

3.22

60)*

0.65

62 (

4.75

44)*

GA

RC

H(1

,1)

4.63

E-0

5 (2

.408

8)*

0.08

85 (

2.43

46)*

0.83

77 (

15.7

03)*

EG

AR

CH

–3.0

397

(-2.

7150

)*0.

2834

(2.

6735

)*–0

.182

4 (-

2.78

99)*

0.62

26 (

4.22

67)*

TG

AR

CH

0.00

01 (

2.72

82)*

0.04

77 (

0.81

96)

0.28

16 (

2.46

40)*

0.50

75 (

3.74

49)*

B.P

ost-

futu

res

perio

d

MA

(1)-

GA

RC

H(1

,1)

7.82

E-0

5 (2

.681

9)*

0.21

08 (

2.79

70)*

0.58

71 (

5.27

39)*

MA

(1)-

EG

AR

CH

–1.8

263

(-2.

6472

)*0.

3521

(3.

1675

)*–0

.076

9 (-

0.91

91)

0.80

28 (

9.70

76)*

GA

RC

H(1

,1)

8.34

E-0

5 (2

.906

1)*

0.23

27 (

2.92

25)*

0.55

86 (

5.12

36)*

EG

AR

CH

–1.8

680

(-2.

8931

)*0.

3813

(3.

4953

)*–0

.077

9 (-

1.02

86)

0.79

99 (

10.2

91)*

TG

AR

CH

7.62

E-0

5 (3

.033

9)*

0.15

30 (

1.95

38)*

0.15

66 (

1.07

82)

0.58

53 (

6.00

02)*

* Si

gnifi

cant

at

the

5 pe

r ce

nt le

vel.

ARCH parameter suggests that good newshas a greater impact. Also, an increase inthe GARCH parameter indicates that oldnews has a greater persistent effect on pricechanges. Notice that similar findings arisefrom the TGARCH model.

The results from other GARCH modelsare presented in Table 5 for the pre-futuresand post-futures period (t-statistics in theparentheses).

Both GARCH(1,1) and MA(1)-GARCH(1,1) reveal similar results. Specifically, allparameters are non-negative (and statisticallysignificant), indicating that GARCH(1,1)models are well specified.38 Thus, therehave been significant changes in thevolatility structure of the FTSE/ASE-20spot market after the introduction of futurestrading. In addition, the evidence indicatesan increase in the ARCH parameter, whichsuggests that news is reflected in pricesmore rapidly. Also, a decrease in theGARCH parameter suggests that old newshas a less persistent effect on price changes.Therefore, old news will have a lowerimpact on today’s price changes. The sumof the coefficients a1 and �1 changes from0.9015 (pre-futures) to 0.7979 (post-futures)for the MA(1)-GARCH(1,1), and from0.9262 (pre-futures) to 0.7913 (post-futures)for the simple GARCH(1,1) model. Hence,the persistence of shocks from thepre-futures period to the post-futures periodis reduced, indicating increased marketefficiency. This is also confirmed by thereduction of the GARCH parameter (�1).Notice that, the difference in the sum ofthe GARCH coefficients is unlikely to beeither statistically or economicallysignificant. It is quite unlikely that 0.9015 is

statistically significant from 0.9262. In thesame line, the difference of the twocoefficients is not economically significant.Nevertheless, the sign of MA(1) coefficientmay be useful in capturingnon-synchronous, infrequent trade in theFTSE-20 index.

From EGARCH(1,1) models, there is anincrease in a1 and a3 parameters. Also, theleverage effect term in bothMA(1)-EGARCH(1,1) and EGARCH(1,1)models is negative. In the pre-futuresperiod, the leverage effect term isstatistically different from zero, indicatingthe existence of leverage in stock returnsduring the sample period. In thepost-futures period, the leverage effect termis not significant.

FTSE/ASE Mid 40 index

Next, estimates of the effect of futurestrading on stock market volatility for theFTSE/ASE Mid 40 index are reported.


Table 6: Results of GARCH models forFTSE/ASE MID 40 index

Coeff. on Model dummy t ratio

AR(3)-TGARCH 4.11E-05 1.7904*

AR(3)-GARCH (1,1) –4.69E-07 –0.0111

AR(3)-EGARCH 0.0461 2.3694*

GARCH (1,1) –3.83E-05 –0.6211

EGARCH –0.0133 –0.2739

TGARCH –1.60E-05 –0.3209


however, decreases (from 1.0449 to0.7761), indicating that old news has a lesspersistent effect on prices and, therefore,old news will have less impact on today’sprice changes.

For the TGARCH model, leverage ispositive but not statistically significant. Also,both ARCH and GARCH parametercoefficients fall. This means good news hasa relatively slow impact with lowpersistence (there is a decrease in a3). Table8 reports the estimation results of variousGARCH models before and after theintroduction of stock index futures.

For GARCH models, all parameters arenon-negative, indicating the superiority ofthe GARCH(1,1) models. In bothAR(3)-GARCH(1,1) and GARCH(1,1),there is a decrease in the ARCHparameter, which suggests that news isbeing reflected in prices slowly. The

Table 6 shows that there is a positive effectof futures trading on stock market volatilitybased on the AR(3)-TGARCH(1,1) andAR(3)-EGARCH(1,1) models. This effectis statistically significant, and there is anincrease in volatility associated with futuresintroduction. The other GARCH modelsindicate a decrease in volatility (ie negativeeffect). This effect is not statisticallysignificant in all cases.

The results from theAR(3)-TGARCH(1,1) model are presentedin Table 7. For this model, the leverageeffect term is positive and statisticallysignificant. So leverage exists, and there isan asymmetric effect to the news. Inaddition, the a1 parameter changes from–0.1797 (pre-futures) to 0.0747(post-futures). Nevertheless, both t ratiosindicate that these parameters areinsignificant. The GARCH parameter (a3),


Table 7: FTSE/ASE MID 40 index AR(3)-TGARCH: dependent variable R

Mean equation Coefficient t-statistic Variance equation

A. Pre-futures period

� –0.0110 –4.6507* � –1.53E-05 –0.9185

AR(1) 0.1501 1.1929 a1 –0.1797 –1.0374

AR(2) –0.3403 –2.7025* a2 0.5088 2.2920*

AR(3) –0.0972 –0.9300 a3 1.0449 6.2609*

B. Post-futures period

� –0.0039 –3.4578* � 4.49E-05 1.9556*

AR(1) 0.1804 3.4773* a1 0.0747 1.3070

AR(2) –0.1454 –2.4790* a2 0.1655 1.6189

AR(3) 0.0709 1.1719 a3 0.7761 10.572*



Tabl

e 8:

Est

imat

ion

resu

lts o

f G

AR

CH

mod

els

for

FTS

E/A

SE

MID

40

inde

x

Mod

els

�a 1

a 2a 3

�1

A.P

re-fu

ture

s pe

riod

AR

(3)-

GA

RC

H(1

,1)

–1.5

7E-0

5 (-

0.36

99)

0.30

65 (

1.87

86)*

0.70

50 (

6.11

71)*

AR

(3)-

EG

AR

CH

–9.2

383

(-2.

8020

)*–0

.578

4 (-

1.22

77)

0.44

84 (

1.81

94)*

–0.4

212

(-0.

8926

)

GA

RC

H(1

,1)

–4.1

4E-0

5 (-

0.76

68)

0.28

27 (

2.05

05)*

0.74

04 (

6.12

31)*

EG

AR

CH

0.51

60 (

1.51

35)

–0.2

515

(-0.

5529

)–0

.372

3 (-

3.42

1512

)*1.

0355

(12

1.33

)*

TG

AR

CH

–6.5

5E-0

5 (-

1.09

70)

0.14

39 (

1.30

37)

0.20

10 (

0.78

89)

0.80

24 (

5.93

25)*

B.P

ost-

futu

res

perio

d

AR

(3)-

GA

RC

H(1

,1)

7.03

E-0

5 (2

.013

2)*

0.16

67 (

2.76

57)*

0.70

89 (

6.74

65)*

AR

(3)-

EG

AR

CH

–0.4

974

(-2.

4167

)*0.

2026

(3.

2929

)*–0

.103

3 (-

1.84

53)*

0.95

42 (

39.5

41)*

GA

RC

H(1

,1)

7.41

E-0

5 (2

.162

6)*

0.18

84 (

2.98

55)*

0.68

82 (

6.86

82)*

EG

AR

CH

–0.9

229

(-2.

5212

)*0.

2936

(3.

4892

)*–0

.094

8 (-

1.52

47)

0.90

66 (

20.8

84)*

TG

AR

CH

5.37

E-0

5 (2

.075

1)*

0.10

09 (

1.73

09)*

0.16

09 (

1.51

08)

0.74

01 (

9.38

72)*

* Si

gnifi

cant

at

the

5 pe

r ce

nt le

vel.

have finite unconditional variances (see alsoBologna and Cavallo1). The unconditionalvariance (� 2) has the form

� 2 ��

1 a1 �1(4)

Comparing parameters across the twosub-periods, it is found that for theFTSE/ASE-20 contract there is an increasein the ARCH parameter, while theGARCH parameter decreases. Now, for theMA(1)-GARCH(1,1) model, theunconditional variance is equal to0.000609651 for the pre-futures period and0.000387044 for the post-futures period. Inaddition, for the simple GARCH(1,1)model, the unconditional variance is equalto 0.000628171 for the pre-futures periodand 0.000399753 for the post-futuresperiod. In other words, the unconditionalvariance in the post-futures period is lowerthan that in the pre-futures period. Thisindicates lower market volatility after theintroduction of stock index futures in theGreek stock market.

For the FTSE/ASE Mid 40 index, theunconditional variance is still lower in thepost-futures period. More specifically, forthe AR(3)-GARCH(1,1) model, theunconditional variance is equal to0.001347408 for the pre-futures periodand 0.000565422 for the post-futuresperiod. Also, for the simple GARCH(1,1)model, the unconditional variance changesfrom 0.001898816 to 0.000600954. Thus,the unconditional variance in thepost-futures period is lower than that ofthe pre-futures period. In other words,the volatility of the Greek stock marketdiminished after the introduction of stock

ARCH coefficient parameters arestatistically significant, but estimates of theGARCH parameters differ. ForAR(3)-GARCH(1,1), there is increase inthe GARCH parameter (from 0.7050 to0.7089), which suggests that old news has agreater persistent effect on price changes.For the simple GARCH(1,1) model, thereduction of the GARCH parametersuggests that old news has a less persistenteffect, and that old news has a lowerimpact on today’s price changes. Also, asfor the FTSE/ASE-20 index, the suma1 � �1 reduces from 1.0116 (pre-futures) to0.8756 (post-futures) for theAR(3)-GARCH(1,1) and from 1.0232(pre-futures) to 0.8766 (post-futures) for thesimple GARCH(1,1) model. Therefore,there is an increase in market efficiency. Inall cases, the coefficient parameters arestatistically significant.

Finally, EGARCH models giveinteresting information about the futurestrading effect. In particular, forAR(3)-EGARCH, the leverage effect termchanges from 0.4484 to –0.1033. For thesimple EGARCH model, the leverageeffect term changes from –0.3723 to–0.0948. Notice that in the pre-futuresperiod, a2 < 0 and significant, indicating thata leverage effect exists in stock returnsduring the sample period. Hence, anegative shock increases the conditionalvariance.

Unconditional variance

In most GARCH (1,1) models, the ARCHand GARCH parameters are non-negative.Also since the sum a1 � 1 for the GARCH(1,1) model is less than one, the models


index futures trading.This finding is in line with what Bologna

and Cavallo1 found for the Italian stockmarket. Finally, the impact of futurestrading on the rate at which information isincorporated into spot prices is illustrated,

and the effect of the shock to both thepre- and post-futures periods is plotted.Figures 1 and 2 plot the effect of the shockfor the FTSE/ASE-20, and Figures 3 and 4plot the effect of the shock for theFTSE/ASE Mid 40. From these figures it is


Figure 1: Effect of a 1 SD shock on spot price volatility before the onset of futurestrading (FTSE/ASE-20)

Figure 2: Effect of a 1 SD shock on spot price volatility after the onset of futures trading(FTSE/ASE-20)

0.000

0.005

0.010

0.015

0.020

0.025

2 4 6 8 10 12 14 16 18 20

–0.005

0.00

0.00

0.01

0.01

0.02

2 4 6 8 10 12 14 16 18 20

in particular, the impact of futures on stockmarket volatility has generated a longdebate. Previous studies have shown thatthe futures market leads to an increase inmarket depth and a decrease in volatility.This is due to the more rapid rate at whichinformation is reflected in prices and

clear that the impact of the shock is morepersistent in the post-futures period than inthe pre-futures period for both indices.

SUMMARY AND CONCLUSIONS

The introduction of a futures market and,


Figure 3: Effect of a 1 SD shock on spot price volatility before the onset of futurestrading (FTSE/ASE Mid 40)

Figure 4: Effect of a 1 SD shock on spot price volatility after the onset of futures trading(FTSE/ASE Mid 40)

–0.01

0.00

0.01

0.02

0.03

0.04

2 4 6 8 10 12 14 16 18 20

–0.005

0.000

0.005

0.010

0.015

0.020

0.025

2 4 6 8 10 12 14 16 18 20

speculation. Other studies suggest that adecrease in cash market volatility is due toan increase in market liquidity. Empiricalstudies for UK and US financial markets donot conclude clearly whether theintroduction of futures stabilises ordestabilises the underlying spot market.

This study has analysed the effect of theimpact of the introduction of futures onother stock prices on the Greek StockExchange. To the authors’ knowledge, thisis the first study that has examined theeffect of the Greek futures market on stockmarket volatility.

A significant indicator of this effect isspot market volatility. To analyse therelationship between stock index futuresand stock market volatility, severalGARCH models were used to model theFTSE/ASE-20, FTSE/ASE Mid 40 andGeneral ASE indexes.

For the FTSE/ASE-20, the results of theeffect of futures trading suggest that therehas been a negative effect on spot pricevolatility during the period. For the subperiods, we find that good news has amore rapid impact on stock returnvolatility, and that the persistence of shocksis reduced, indicating increased market(pricing) efficiency. This is not surprising, asthe FTSE/ASE-20 futures market is highlyliquid. In addition, the results suggest thatold news has either a greater or aless-persistent effect on price changes. Thefact that noise traders39 are the principalusers of the FTSE/ASE-20 contractindicates that volatility becomes morepersistent, so an increase in persistence isnot surprising. Further, GARCH(1,1)models show that futures trading improves

the speed of information flow to the spotmarket. This is in line with arguments fromAntoniou and Holmes7 for the FTSE 100index.

For the FTSE/ASE Mid 40 index, theempirical results are mixed. AsymmetricAR(3)-EGARCH and AR(3)-TGARCHmodels show a positive effect on pricevolatility during the period examined.GARCH-type models (where theconditional mean equation just includes aconstant term) show a negative (but notsignificant) effect, however. Furthermore,there has been a decrease in both theARCH and GARCH parameters, indicatingthat news is being reflected in prices moreslowly, and that old news has a lesspersistent effect on prices. The decrease inthe coefficient on past variance shows that,after the onset of futures trading, spotmarket volatility is not important to spotmarket participants (see Antoniou andFoster18). The main reason is that the onsetprice risk can be hedged in the futuresmarket. Further, the TGARCH andEGARCH specifications show that theleverage effect exists and, thus, there is anasymmetric effect to the news. Also, theEGARCH model shows that a negativeshock increases the conditional variance.

In summary, the evidence suggests anegative effect of futures trading on Greekstock market volatility (however, this is notvery strong for FTSE/ASE Mid 40 index).This is confirmed by the estimation ofthree different types of GARCHspecifications and unconditional variances.In particular, the unconditional variance inthe post-futures period is found to be lowerthan that of the pre-futures period. This


& Accounting, Vol. 18, pp. 267–280.9 Edwards, F. R. (1988) ‘Futures Trading and Cash

Market Volatility: Stock Index and Interest RateFutures’, Journal of Futures Markets, Vol. 8, pp.421–439.

10 Chang, E. C., Cheng, J. and Pinegar, J. M.(1999) ‘Does Futures Trading Increase StockMarket Volatility? The Case of the Nikkei StockIndex Futures Markets’, Journal of Banking andFinance, Vol. 23, pp. 727–753.

11 Aggarwal, R. (1988) ‘Stock Index Futures andCash Market Volatility’, Review of FuturesMarkets, Vol. 7, No. 2, pp. 290–299.

12 Harris, L. (1989) ‘S&P500 Cash Stock PriceVolatilities’, Journal of Finance, Vol. 44,pp. 1155–1175.

13 Lockwood, L. J., and Linn, S. C. (1990) ‘AnExamination of Stock Market Return VolatilityDuring Overnight and Intraday Periods1964–1989’, Journal of Finance, Vol. 45,pp. 591–601.

14 Maberly, E. D., David, S. A. and Roy, R. G.(1989) ‘Stock Index Futures and Cash MarketVolatility’, Financial Analysts Journal, Vol. 45,pp. 75–77.

15 Chang, E., Jain, P. and Locke, P. (1995) ‘Standard& Poor’s 500 Index Futures Volatility and PriceChanges Around the New York Stock ExchangeClose’, Journal of Business, Vol. 68, pp. 61–84.

16 Chang, E., Chou, R. and Nelling, E. (2000)‘Market Volatility and the Demand for Hedgingin Stock Index Futures’, Journal of Futures Markets,Vol. 20, pp. 105–125.

17 Chan, K., Chan, K. C., and Karolyi, G. A.(1991) ‘Intraday Volatility in the Stock Index andStock Index Futures Markets, Review of FinancialStudies, Vol. 4, pp. 657–684.

18 Antoniou, A. and Foster, A. J. (1992) ‘The Effectof Futures Trading on Spot Price Volatility:Evidence for Brent Crude Oil Using GARCH’,Journal of Business Finance & Accounting, Vol. 19,No. 4, pp. 473–484.

19 Brorsen, B. W. (1991) ‘Futures Trading,Transaction Costs, and Stock Market Volatility’,Journal of Futures Markets, Vol. 11, pp. 153–163.

20 Gulen, H. and Mayhew, S. (2000) ‘Stock IndexFutures Trading and Volatility in InternationalEquity Markets’, Journal of Futures Markets,Vol. 20, pp. 661–685.

21 Holmes, P. (1996) ‘Spot Price Volatility,Information and Futures Trading: Evidence froma Thinly Traded Market’, Applied EconomicsLetters, Vol. 3, pp. 63–66.

22 Robinson, G. (1993) ‘The Effects of Futures

indicates lower market volatility after theintroduction of stock index futures. This isconsistent with Bologna and Cavallo1 forthe Italian Stock Exchange.

Finally, future research should test (1)whether the introduction of stock indexfutures affects the volume-volatilityrelationship in the spot market, usingGARCH and other stochastic volatilitymodels, and (2) for the presence of causalrelationships, using VAR representation andGranger causality tests.

References and Notes

1 Bologna, P. and Cavallo, L. (2002) ‘Does theIntroduction of Stock Index Futures EffectivelyReduce Stock Market Volatility? Is the ‘FuturesEffect’ Immediate? Evidence from the ItalianStock Exchange using GARCH’, Applied FinancialEconomics, Vol. 12, pp. 183–192.

2 S. Kyle (1985 ‘Continuous Auctions and InsiderTrading’, Econometrica, Vol. 53, pp. 1315–1335)defines market depth as the order flow requiredto move prices by one unit. Market depth isrelated to non-informational trading activity andprovides additional information about theinteraction between price volatility and tradingvolume.

3 Stein, J. (1987) ‘Informational Externalities andWelfare-reducing Speculation’, Journal of PoliticalEconomy, Vol. 95, pp. 1123–1145.

4 Subrahmanyam, A. (1991) ‘A Theory of Tradingin Stock Index Futures’, Review of FinancialStudies, Vol. 4, pp. 17–51.

5 Newbery, D. M. (1987) ‘When Do FuturesDestabilize Spot Prices?’, International EconomicReview, Vol. 28, pp. 291–297.

6 Chari, V. V., Jagannathan, R. and Jones, L.(1990) ‘Price stability and futures trading incommodities’, Quarterly Journal of Economics, Vol.105, pp. 527–34.

7 Antoniou, A. and Holmes, P. (1995) ‘FuturesTrading, Information and Spot Price Volatility:Evidence for the FTSE 100 Stock Index FuturesContract using GARCH’, Journal of Banking &Finance, Vol. 19, pp. 117–129.

8 Hodgson, A. and Nicholls, D. (1991) ‘TheImpact of Index Futures Markets on AustralianSharemarket Volatility’, Journal of Business Finance


Trading on Cash Market Volatility: Evidencefrom the London Stock Exchange’, Bank ofEngland, Working Paper, Series No 19.

23 Antoniou, A., Holmes, P. and Priestley, R.(1998) ‘The Effects of Stock Index FuturesTrading on Stock Index Volatility: An Analysis ofthe Asymmetric Response of Volatility to News,Journal of Futures Markets, Vol. 18, pp. 151–166.

24 Butterworth, D. (1998) ‘The Impact of FuturesTrading on Underlying Stock Index Volatility:The Case of the FTSE Mid 250 Contract’,Working Paper, Department of Economics,University of Durham.

25 Rahman, S. (2001) ‘The introduction ofderivatives on the Dow Jones Industrial Averageand their impact on the volatility of componentstocks’, Journal of Futures Markets, Vol. 21,pp. 633–653.

26 Koutmos, G. and Tucker, M. (1996) ‘TemporalRelationships and Dynamic Interactions BetweenSpot and Futures Stock Markets’, Journal ofFutures Markets, Vol. 16, pp. 55–69.

27 Pericli, A. and Koutmos, G. (1997) ‘IndexFutures and Options and Stock Market Volatility’,Journal of Futures Markets, Vol. 17, pp. 957–974.

28 Darrat, A. F., Rahman, S. and Zhong, M. (2000)‘On the role of futures trading in sport marketfluctuations: Perpetrator of volatility or victim ofregret?’, Paper Presented at FinancialManagement Association International 2000Annual Meeting, Seattle, Washington.

29 Lien, D. and Tse, Y. K. (1999) ‘The Effects ofCash Settlement on the Cash-Futures Prices andTheir Relationship: Evidence from the FeederCattle Contract’, Working Paper, University ofKansas.

30 Kyriacou, K. and Sarno, L. (1999) ‘The TemporalRelationship between Derivatives Trading andSpot market Volatility in the U. K.: EmpiricalAnalysis and Monte Carlo Evidence’, Journal ofFutures Markets, Vol. 19, pp. 245–270.

31 Pilar, C. and Rafael, S. (2002) ‘Does DerivativesTrading Destabilize the Underlying Assets?Evidence from the Spanish Stock Market’,Applied Economics Letters, Vol. 9, pp. 107–110.

32 Figuerola-Ferretti, I. and Gilbert, C. L. (2001)‘Has Futures Trading Affected the Volatility ofAluminium Transaction Prices?’, Working Paper,Department of Economics, Queen Mary,University of London.

33 Chiang, M.-H. and Wang, C.-Y. (2002) ‘TheImpact of Futures Trading on Spot IndexVolatility: Evidence for Taiwan Index Futures’,Applied Economics Letters, Vol. 9, pp. 381–385.

34 Bae, S. C., Kwon, T. H. and Park, J. W. (2004)‘Futures Trading, Spot Market Volatility, andMarket Efficiency: the Case of the Korean IndexFutures Markets’, Journal of Futures Markets,Vol. 24, pp. 1195–1228.

35 Ryoo, H.-J. and Smith, G. (2004) ‘The Impactof Stock Index Futures on the Korean StockMarket’, Applied Financial Economics, Vol. 14,pp. 243–251.

36 Antoniou, A., Koutmos, G. and Pericli, A. (2005)‘Index Futures and Positive Feedback Trading:Evidence from Major Stock Exchanges’, Journal ofEmpirical Finance, Vol. 12, pp. 219–238.

37 Floros, C. and Vougas, D. V. (2004) ‘HedgeRatios in Greek Stock Index Futures Market’,Applied Financial Economics, Vol. 14, No. 15,pp. 1125–1136.

38 The GARCH(1,1) model has been found to bethe most parsimonious representation ofconditional variance that best fits many financialseries, see Bollerslev,39 Butterworth,24 Antoniouand Holmes7 and Bologna and Cavallo.1

39 Bollerslev, T. (1987) ‘A ConditionalHeteroscedastic Time Series Model forSpeculative Prices and Rates of Return’, Reviewof Economics and Statistics, Vol. 69, pp. 542–547.

40 A noise trader is a trader who trades for pleasureor who trades on information that he or shebelieves is valuable but which is not useful.Noise traders can be treated as investors who donot use fundamental information or whootherwise fail to use it ‘fully and correctly’. G.W. Schwert (1990 ‘Stock Volatility and the Crashof ‘87’, Review of Financial Studies, Vol. 3,pp. 77–102) suggests that noise traders are moreactive in futures than in spot markets.



APPENDIX 1

A. GARCH variance series before and after the introduction of the FTSE/ASE-20

stock index futures

0.0000

0.0005

0.0010

0.0015

0.0020

0.0025

0.0030

0.0035

50 100 150 200 250 300 350 400 450 500

Garch variance before

0.0000

0.0005

0.0010

0.0015

0.0020

0.0025

0.0030

50 100 150 200 250 300 350 400 450 500

Garch variance after


B. GARCH variance series before and after the introduction of the FTSE/ASE Mid

40 stock index futures

0.000

0.001

0.002

0.003

0.004

0.005

0.006

5 10 15 20 25 30 35

Garch variance before

0.0000

0.0005

0.0010

0.0015

0.0020

0.0025

0.0030

0.0035

50 100 150 200 250 300 350

Garch variance after

Documents

Index futures trading, information and stock market ... · futures trading activity on the volatility of the FTSE 100 index. Antoniou et al.,23 however, examine further the impact