Transmission of prices and price volailiy in Ausralian elecriciy spo markes: A mulivariae GARCH analysis Auhor Worhingon, Andrew, Kay-Spraley, Adam, Higgs, Helen Published 2005 Journal Tile Energy Economics DOI hps://doi.org/10.1016/j.eneco.2003.11.002 Copyrigh Saemen 2005 Elsevier. This is he auhor-manuscrip version of his paper. Reproduced in accordance wih he copyrigh policy of he publisher. Please refer o he journal's websie for access o he definiive, published version. Downloaded from hp://hdl.handle.ne/10072/16241 Griffih Research Online hps://research-reposiory.griffih.edu.au
Transmission of prices and price volailiy in Ausralian elecriciy spo markes: A mulivariae GARCH analysis Andrew Worhingon *, Adam Kay-Spraley, Helen Higgs School of Economics and Finance, Queensland Universiy of Technology, GPO Box 2434, Brisbane, QLD 4001, Ausralia Absrac This paper examines he ransmission of spo elecriciy prices and price volailiy among he five regional elecriciy markes in he Ausralian Naional Elecriciy Marke (NEM): namely, New Souh Wales (NSW), Queensland (QLD), Souh Ausralia (SA), he Snowy Mounains Hydroelecric Scheme (SNO) and Vicoria (VIC). A mulivariae generalised auoregressive condiional heeroskedasiciy (MGARCH) model is used o idenify he source and magniude of price and price volailiy spillovers. The resuls indicae he presence of posiive own mean spillovers in only a small number of markes and no mean spillovers beween any of he markes. This appears o be direcly relaed o he physical ransfer limiaions of he presen sysem of regional inerconnecion. Neverheless, he large number of significan own-volailiy and cross-volailiy spillovers in all five markes indicaes he presence of srong ARCH and GARCH effecs. This indicaes ha shocks in some markes will affec price volailiy in ohers. Finally, and conrary o evidence from sudies in Norh American elecriciy markes, he resuls also indicae ha Ausralian elecriciy spo prices are saionary. JEL classificaions: C32. C51, L94, Q40 Keywords: spo elecriciy price markes; mean and volailiy spillovers; mulivariae GARCH 1. Inroducion The Ausralian Naional Elecriciy Marke (NEM) was esablished on 13 December 1998. I currenly comprises four sae-based [New Souh Wales, Vicoria, Queensland and Souh Ausralia] and one non-sae based [Snowy Mounains Hydroelecric Scheme] regional markes operaing as a naionally inerconneced grid. Wihin his grid, he larges generaion capaciy is found in New Souh Wales, followed by Queensland, Vicoria, he Snowy Mounains Hydroelecric Scheme and Souh Ausralia, while elecriciy demand is highes in New Souh Wales, followed by Vicoria, Queensland and Souh Ausralia. The more han seveny regisered paricipans in he NEM, encompassing privaely and publicly owned generaors, ransmission and disribuion nework providers and raders, currenly supply * Corresponding auhor. Tel: +61-7-3864-2658; fax: 61-703864-1500. E-mail: a.worhingon@qu.edu.au.
2 A. Worhingon, A. Kay-Spraley, H. Higgs elecriciy o 7.7 million cusomers wih more han $8 billion of energy raded annually [for deails of he NEM s regulaory background, insiuions and operaions see NEMMCO (2001, 2002), ACCC (2000) and IEA (2001)]. Hisorically, he very gradual move o an inegraed naional sysem was predaed by subsanial reforms on a sae-by-sae basis, including he unbundling of generaion, ransmission and disribuion and he commercializaion and privaizaion of he new elecriciy companies, along wih he esablishmen of he wholesale elecriciy spo markes (Dickson and Warr, 2000). Each sae in he NEM iniially developed is own generaion, ransmission and disribuion nework and linked i o anoher sae's sysem via inerconnecor ransmission lines. However, each sae s nework was (and sill is) characerised by a very small number of paricipans and sizeable differences in elecriciy prices were found. The foremos objecive in esablishing he NEM was hen o provide a naionally inegraed and efficien elecriciy marke, wih a view o limiing he marke power of generaors in he separae regional markes [for he analysis of marke power in elecriciy markes see Brennan and Melanie (1998), Joskow and Kahn (2002), Wilson (2002) and Robinson and Baniak (2002)]. However, a defining characerisic of he NEM is he limiaions of physical ransfer capaciy. Queensland has wo inerconnecors ha ogeher can impor and expor o and from NSW, NSW can expor o and from he Snowy and Vicoria can impor from he Snowy and Souh Ausralia and expor o he Snowy and o Souh Ausralia. There is currenly no direc connecor beween NSW and Souh Ausralia (hough one is proposed) and Queensland is only direcly conneced o NSW. As a resul, he NEM iself is no ye srongly inegraed wih inersae rade represening jus seven percen of oal generaion. During periods of peak demand, he inerconnecors become congesed and he NEM separaes ino is regions, promoing price differences across markes and exacerbaing reliabiliy problems and he marke power of regional uiliies (IEA, 2001; ACCC, 2000; NEMMCO, 2002). While he appropriae regulaory and commercial mechanisms do exis for he creaion of an efficien naional marke, and hese are expeced o have an impac on he price of elecriciy in each jurisdicion, i is argued ha he complee inegraion of he separae regional elecriciy markes has no ye been realised. In paricular, he limiaions of he inerconnecors beween he member jurisdicions sugges ha, for he mos par, he regional spo markes are relaively isolaed. Neverheless, he Vicorian elecriciy crisis of February 2000 is jus one of several shocks in he Ausralian marke ha suggess spo elecriciy pricing and volailiy in each regional marke are sill poenially dependen on pricing
A. Worhingon, A. Kay-Spraley, H. Higgs 3 condiions in oher markes. These are, of course, concerns ha are likely o be jus as imporan in any oher naional or sub-naional elecriciy marke comprised of inerconneced regions. In he Unied Saes, for example, De Vany and Walls (1999a) used coinegraion analysis o es for price convergence in regional markes in he US Wesern Elecriciy Grid. On he whole he findings were suggesive of an efficien and sable wholesale power marke, hough De Vany and Walls (1999a) argued ha he lack of coinegraion in some markes provided evidence of he impac of ransfer consrains wihin he grid. Laer, De Vany and Walls used vecor auoregressive modelling echniques and variance decomposiion analysis o examine a smaller se of hese regional markes. They concluded he efficiency of power pricing on he wesern ransmission grid is esimony o he abiliy of decenralised markes and local arbirage o produce a global paern of nearly uniform prices over a complex and decenralised ransmission nework spanning vas disances (De Vany and Walls 1999b: 139). Unforunaely, no comparable evidence exiss concerning he inerconneced regional elecriciy markes in Ausralia, or indeed elsewhere ouside he Unied Saes for ha maer. This is imporan for wo reasons. Firs, unlike he US he Ausralian NEM represens he polar case of a cenrally coordinaed and regulaed naional marke. I is herefore likely o hrow ligh on he efficiency of pricing and he impac of inerconnecion wihin cenralised markes sill primarily composed of commercialised and corporaised public secor eniies. Second, a fuller undersanding of he pricing relaionships beween hese markes will enable he benefis of inerconnecion o be assessed as a sep owards he fuller inegraion of he regional elecriciy markes ino a naional elecriciy marke. This provides policy inpus ino boh he consrucion of new inerconnecors and guidelines for he reform of exising marke mechanisms. A he same ime, he manner in which volailiy shocks in regional elecriciy markes are ransmied across ime arouses ineres in modelling he dynamics of he price volailiy process. This calls for he applicaion of auoregressive condiional heeroskedasiciy (ARCH) and generalised ARCH (GARCH) models ha ake ino accoun he ime-varying variances of ime series daa [suiable surveys of ARCH modelling may be found in Bollerslev, e al. (1992), Bera and Higgins (1993) and Pagan (1996)]. More recenly, he univariae GARCH model has been exended o he mulivariae GARCH (MGARCH) case, wih he recogniion ha MGARCH models are poenially useful developmens regarding he parameerizaion of condiional cross-momens. Alhough he MGARCH mehodology has
4 A. Worhingon, A. Kay-Spraley, H. Higgs been used exensively in modelling financial ime series [see, for insance, Dunne (1999), Tai (2000), Brooks e al. (2002) and Tse and Tsui (2002)], o he auhors knowledge a deailed sudy of he applicaion of MGARCH o elecriciy markes has no been underaken. Since his approach capures he effec on curren volailiy of boh own innovaion and lagged volailiy shocks emanaing from wihin a given marke and cross innovaion and volailiy spillovers from inerconneced markes i permis a greaer undersanding of volailiy and volailiy persisence in hese inerconneced markes. I is wihin he conex of his limied empirical work ha he presen sudy is underaken. Accordingly, he purpose of his paper is o invesigae he price and price volailiy inerrelaionships beween he Ausralian regional elecriciy markes. If here is a lack of significan inerrelaionships beween regions hen doub may hen be cas on he abiliy of he NEM o overcome he exercise of regional marke power as is primary objecive, and on is capaciy o foser a naionally inegraed and efficien elecriciy marke. The paper iself is divided ino four secions. The second secion explains he daa employed in he analysis and presens some brief summary saisics. The hird secion discusses he mehodology employed. The resuls are deal wih in he fourh secion. The paper ends wih some brief concluding remarks. 2. Daa and summary saisics The daa employed in he sudy are daily spo prices for elecriciy encompassing he period from he dae of commencemen of he Naional Elecriciy Marke (NEM) on 13 December 1998 o 30 June 2001. The sample period is chosen on he basis ha i represens a coninuous series of daa since he esablishmen of he Ausralian Naional Elecriciy Marke (NEM). All price daa is obained from he Naional Elecriciy Marke Managemen Company (NEMMCO) originally on a half-hourly basis represening 48 rading inervals in each 24-hour period. Following Lucia and Schwarz (2001) a series of daily arihmeic means is drawn from he rading inerval daa. Alhough such reamen enails he loss of a leas some news impounded in he more frequen rading inerval daa, daily averages play an imporan role in elecriciy markes, paricularly in he case of financial conracs. For example, he elecriciy srips raded on he Sydney Fuures Exchange (SFE) are seled agains he arihmeic mean of half hourly spo prices. Moreover, De Vany and Walls (1999a; 1999b) and Robinson (2000) boh employ daily spo prices in heir respecive analyses of he wesern Unied Saes and Unied Kingdom spo elecriciy markes.
A. Worhingon, A. Kay-Spraley, H. Higgs 5 <TABLE 1 HERE> Table 1 presens he summary of descripive saisics of he daily spo prices for he five elecriciy markes. Samples means, medians, maximums, minimums, sandard deviaions, skewness, kurosis and he Jacque-Bera saisic and p-value are repored. Beween 13 December 1998 and 30 June 2001, he highes spo prices are in Queensland (QLD) and Souh Ausralia (SA) averaging $42.71 and $57.92 per megawa-hour, respecively. The lowes spo prices are in New Souh Wales (NSW) and he Snowy Mounains Hydroelecric Scheme (SNO) wih $33.02 and $32.56, respecively. The sandard deviaions for he spo elecriciy range from $27.84 (Snowy Mounains Hydroelecric Scheme) o $92.15 (Souh Ausralia). Of he five markes, New Souh Wales (NSW) and he Snowy Mounains Hydroelecric Scheme (SNO) are he leas volaile, while Queensland (QLD) and Souh Ausralia (SA) are he mos volaile. The value of he of variaion (sandard deviaion divided by he mean price) measures he degree of variaion in spo price relaive o he mean spo price. Relaive o he average spo price, New Souh Wales (NSW) and he Snowy Mounains Hydroelecric Scheme (SNO) are less variable han Souh Ausralia (SA) and Vicoria (VIC). A visual perspecive on he volailiy of spo prices can be gained from he plos of daily spo prices for each series in Figure 1. <FIGURE 1 HERE> The disribuional properies of he spo price series generally appear non-normal. All of he spo elecriciy markes are posiively skewed and since he kurosis, or degree of excess, in all of hese elecriciy markes exceeds hree, a lepokuric disribuion is indicaed. The calculaed Jarque-Bera saisic and corresponding p-value in Table 1 is used o es he null hypoheses ha he daily disribuion of spo prices is normally disribued. All p-values are smaller han he.01 level of significance suggesing he null hypohesis can be rejeced. These daily spo prices are hen no well approximaed by he normal disribuion. Lasly, each price series is esed for he presence of a uni roo using he Augmened Dickey-Fuller (ADF) es. Conrary o previous empirical work De Vany and Walls (1999a; 1999b), which found ha spo elecriciy prices conain a uni roo, his sudy concurs wih Lucia and Schwarz (2001) ha elecriciy prices are saionary.
6 A. Worhingon, A. Kay-Spraley, H. Higgs 3. Mehodology A MGARCH model is developed o examine he join processes relaing he daily spo prices for he five regional elecriciy markes. The following condiional expeced price equaion accommodaes each marke s own prices and he prices of oher markes lagged one period. P + ε = α + AP 1 where P is an n 1 vecor of daily prices a ime for each marke and ε I ~N ( 0, H 1 ). The n 1 vecor of random s, ε is he innovaion for each marke a ime wih is corresponding n n condiional variance-covariance marix, H. The marke informaion available a ime - 1 is represened by he informaion se I -1. The n 1 vecor, α, represen long-erm drif s. The elemens a ij of he marix A are he degree of mean spillover effec across markes, or pu differenly, he curren prices in marke i ha can be used o predic fuure prices (one day in advance) in marke j. The esimaes of he elemens of he marix, A, can provide measures of he significance of he own and cross-mean spillovers. This mulivariae srucure hen enables he measuremen of he effecs of he innovaions in he mean spo prices of one series on is own lagged prices and hose of he lagged prices of oher markes. Engle and Kroner (1995) presen various MGARCH models wih variaions o he condiional variance-covariance marix of equaions. For he purposes of he following analysis, he BEKK (Baba, Engle, Kraf and Kroner) model is employed, whereby he variance-covariance marix of equaions depends on he squares and cross producs of innovaion ε and volailiy H for each marke lagged one period. One imporan feaure of his specificaion is ha i builds in sufficien generaliy, allowing he condiional variances and covariances of he elecriciy markes o influence each oher, and, a he same ime, does no require he esimaion of a large number of parameers (Karolyi 1995). The model also ensures he condiion of a posiive semi-definie condiional variance-covariance marix in he opimisaion process, and is a necessary condiion for he esimaed variances o be zero or posiive. The BEKK parameerisaion for he MGARCH model is wrien as: H = B B + C ε ε 1C + G H 1 G where b ij are elemens of an n n symmeric marix of consans B, he elemens c ij of he symmeric n n marix C measure he degree of innovaion from marke i o marke j, and he -
A. Worhingon, A. Kay-Spraley, H. Higgs 7 elemens g ij of he symmeric n n marix G indicae he persisence in condiional volailiy beween marke i and marke j. This can be expressed for he bivariae case of he BEKK as: H H 11 21 H H 12 22 c c = B B + c21c 11 12 22 ε ε ε 2 1 1 1 1 ε ε 2 1 1 1 2 1 ε 2 2 1 c c c21c 11 12 22 + g g 11 21 g g 12 22 H H In his parameerizaion, he parameers b ij, c ij and g ij canno be inerpreed on an individual basis: insead, he funcions of he parameers which form he inercep erms and he s of he lagged variance, covariance, and erms ha appear are of ineres (Kearney and Paon 2000: 36). Wih he assumpion ha he random s are normally disribued, he log-likelihood funcion for he MGARCH model is: L T Tn 1 ' 1 ( θ) = + ln( 2π) ( ln H + ε H ε ) 2 2 = 1 where T is he number of observaions, n is he number of markes, θ is he vecor of parameers o be esimaed, and all oher variables are as previously defined. The BHHH (Bernd, Hall, Hall and Hausman) algorihm is used o produce he maximum likelihood parameer esimaes and heir corresponding asympoic sandard s. Overall, he proposed model has weny-five parameers in he mean equaions, excluding he five consan (inercep) parameers, and weny-five inercep, weny-five whie noise and weny-five volailiy parameers in he esimaion of he covariance process, giving one hundred and five parameers in oal. Lasly, he Ljung-Box Q saisic is used o es for independence of higher relaionships as manifesed in volailiy clusering by he MGARCH model [Huang and Yang 2000:329]. This saisic is given by: Q = T p 1 2 ( T + ) ( T j) r ( 2 j) j= 1 where r(j) is he sample auocorrelaion a lag j calculaed from he noise erms and T is he number of observaions. Q is asympoically disribued as χ 2 wih (p - k) degrees of freedom and k is he number of explanaory variables. This es saisic is used o es he null hypohesis ha he model is independen of he higher order volailiy relaionships. 11 1 21 1 H H 12 1 22 1 g g 11 21 g g 12 22
8 A. Worhingon, A. Kay-Spraley, H. Higgs 4. Empirical resuls The esimaed s and sandard s for he condiional mean price equaions are presened in Table 2. All esimaions are made using he S-PLUS saisical sofware wih he GARCH add-on module. For he five elecriciy spo markes only QLD and SNO exhibi a significan own mean spillover from heir own lagged elecriciy price. In boh cases, he mean spillovers are posiive. For example, in QLD a $1.00 per megawa-hour increase in is own spo price will Granger cause an increase of $0.51 per megawa-hour in is price over he nex day. Likewise, a $1.00 per megawa-hour increase in he SNO lagged spo price will Granger cause a $0.70 increase he nex day. Imporanly, here are no significan lagged mean spillovers from any of he spo markes o any of he oher markes. This indicaes ha on average shor-run price changes in any of he five Ausralian spo markes are no associaed wih price changes in any of he oher spo elecriciy markes, despie he conneciviy offered by he NEM. <TABLE 2 HERE> The condiional variance covariance equaions incorporaed in he paper s mulivariae GARCH mehodology effecively capure he volailiy and cross volailiy spillovers among he five spo elecriciy markes. These have no been considered by previous sudies. Table 3 presens he esimaed s for he variance covariance marix of equaions. These quanify he effecs of he lagged own and cross innovaions and lagged own and cross volailiy persisence on he own and cross volailiy of he elecriciy markes. The s of he variance covariance equaions are generally significan for own and cross innovaions and significan for own and cross volailiy spillovers o he individual prices for all elecriciy markes, indicaing he presence of srong ARCH and GARCH effecs. In evidence, 68 percen (seveneen ou of weny-five) of he esimaed ARCH s and 84 percen (weny-one ou of weny-five) of he esimaed GARCH s are significan a he.10 level or lower. <TABLE 3 HERE> Own-innovaion spillovers in all he elecriciy markes are large and significan indicaing he presence of srong ARCH effecs. The own-innovaion spillover effecs range from 0.0915 in VIC o 0.1046 in SNO. In erms of cross-innovaion effecs in he elecriciy markes, pas innovaions in mos markes exer an influence on he remaining elecriciy markes. For example, in he case of VIC cross innovaion in he NSW, SA and SNO markes are
A. Worhingon, A. Kay-Spraley, H. Higgs 9 significan, of which NSW has he larges effec. The excepion o he presence of srong cross innovaion effecs is QLD. No cross innovaions ouside of QLD influence ha marke, and he QLD marke does influence any of he oher elecriciy markes, a leas over he period in quesion. This is consisen wih he role of QLD in he NEM in ha i has only limied direc conneciviy wih jus one oher regional marke (NSW). In he GARCH se of parameers, eighy-four percen of he esimaed s are significan. For NSW he lagged volailiy spillover effecs range from 0.7839 for SA o 0.8412 for QLD. This means ha he pas volailiy shocks in QLD have a greaer effec on he fuure NSW volailiy over ime han he pas volailiy shocks in oher spo markes. Conversely, in QLD he pos volailiy shocks range from 0.65212 for SA o 0.8413 for SNO. In erms of cross-volailiy for he GARCH parameers, he mos influenial markes would appear o be NSW and SNO. Tha is, pas volailiy shocks in he NSW and SNO elecriciy spo markes have he greaes effec on he fuure volailiy in he hree remaining elecriciy markes. The sum of he ARCH and GARCH s measures he overall persisence in each marke s own and cross condiional volailiy. All five elecriciy markes exhibi srong own persisence volailiy ranging from 0.9032 for NSW o 0.9143 for SNO. Thus, SNO has a lead-persisence volailiy spillover effec on he remaining elecriciy markes. The crossvolailiy persisence spillover effecs range from 0.7751 for SA 0.9409 for QLD. <TABLE 4 HERE> Finally, he Ljung-Box (LB) Q saisics for he sandardised residuals in Table 4 reveal ha all elecriciy spo markes are highly significan (all have p-values of less han.01) wih he excepion of SNO (a p-value of 0.1166). Significance of he Ljung-Box (LB) Q saisics for he elecriciy spo price series indicaes linear dependences due o he srong condiional heeroskedasiciy. These Ljung-Box saisics sugges a srong linear dependence in four ou of he five elecriciy spo markes esimaed by he MGARCH model. 5. Conclusions and policy implicaions This paper highlighs he ransmission of prices and price volailiy among five Ausralian elecriciy spo markes during he period 1998 o 2001. All of hese spo markes are member jurisdicions of he recenly esablished Naional Elecriciy Marke (NEM). A he ouse, uni roo ess confirm ha Ausralian elecriciy spo prices are saionary. A mulivariae generalised auoregressive condiional heeroskedasiciy (MGARCH) model is hen used o idenify he source and magniude of spillovers. The esimaed s from he
10 A. Worhingon, A. Kay-Spraley, H. Higgs condiional mean price equaions indicae ha despie he presence of a naional marke for elecriciy, he regional elecriciy spo markes are no inegraed. In fac, only wo of he five markes exhibi a significan own mean spillover. This also would sugges, for he mos par, ha Ausralian spo elecriciy prices could no be usefully forecased using lagged price informaion from eiher each marke iself or from oher markes in he naional marke. However, own-volailiy and cross-volailiy spillovers are significan for nearly all markes, indicaing he presence of srong ARCH and GARCH effecs. Convenionally, his is used o indicae ha markes are no efficien. Srong own and cross-persisen volailiy are also eviden in all Ausralian elecriciy markes. This indicaes ha while he limied naure of he inerconnecors beween he separae regional markes prevens full inegraion, shocks or innovaions in paricular markes sill exer an influence on price volailiy. Thus, during periods of abnormally high demand for example, he NEM may be a leas parially offseing he abiliy of regional paricipans o exer marke power. Noneheless, he resuls mainly indicae he inabiliy of he exising nework of inerconnecors o creae a subsanially inegraed naional elecriciy marke and ha, for he mos par, he sizeable differences in spo prices beween mos of he regions will remain, a leas in he shor erm. This provides validaion for new regional inerconnecors currenly under consrucion and hose ha are proposed, and he anicipaed inclusion of Tasmania as a sixh region in he NEM. As a general rule, he less direc he inerconnecion beween regions, he less significan he cross-innovaion and volailiy spillover effecs beween hese regions. This suggess ha main deerminan of he ineracion beween regional elecriciy markes is geographical proximiy and he number and size of inerconnecors. Accordingly, i may be unreasonable o expec ha prices in elecriciy markes ha are geographically isolaed marke will ever become fully inegraed wih core or geographically proximae markes. The resuls also indicae ha volailiy innovaions or shocks in all markes persis over ime and ha in all markes his persisence is more marked for own-innovaions or shocks han cross-innovaions or shocks. This persisence capures he propensiy of price changes of like magniude o cluser in ime and explains, a leas in par, he nonnormaliy and nonsabiliy of Ausralian elecriciy spo prices. Togeher, hese indicae ha neiher he NEM nor he regional markes are efficienly pricing elecriciy and ha changes o he marke mechanism may be necessary. I may also reinforce calls for he privaisaion of some elecriciy marke paricipans o improve compeiion, given ha he overwhelming majoriy of hese remain under public secor conrol.
A. Worhingon, A. Kay-Spraley, H. Higgs 11 Of course, he full naure of he price and volailiy inerrelaionships beween hese separae markes could be eiher under or oversaed by misspecificaion in he daa, all of which sugges fuure avenues for research. One possibiliy is ha by averaging he half-hourly prices hroughou he day, he speed a which innovaions in one marke influence anoher could be undersaed. For insance, wih he daa as specified he mos rapid innovaion allowed in his sudy is a day, whereas in realiy innovaions in some markes may affec ohers wihin jus a few hours. Similarly, here has been no aemp o separae he differing condiions expeced beween peak and off-peak prices. For example, De Vany and Walls (1999) found ha here were essenially no price differenials beween rading poins in offpeak periods because hey were less consrained by limiaions in he ransmission sysem. Anoher possibiliy is ha he occurrence of ime-dependen condiional heeroskedasiciy could be due o an increased volume of rading and/or variabiliy of prices following he arrival of new informaion ino he marke. I is well known ha financial markes, for insance, can sill be efficien bu exhibi GARCH effecs in price changes if informaion arrives a uneven inervals. One fuure applicaion of modelling would hen include, say, demand volume as a measure of he amoun of informaion ha flows ino he elecriciy marke. This would provide definiive proof of wheher he GARCH effecs are really evidence of marke inefficiency, or he resul of he irregular flow of marke informaion. Research ino Ausralian elecriciy markes could be exended in a number of oher ways. One useful exension would be o examine each of he five elecriciy markes individually and in more deail. For example, while he sample for his sudy is deermined by he period of enure of he Naional Elecriciy Marke (NEM) wholesale elecriciy spo markes in he separae regions pre-dae his by several years. An examinaion of he connecion beween he long-sanding elecriciy spo markes in NSW and Vicoria would be paricularly useful. Anoher suggesion concerns he elecriciy srip conracs offered by he Sydney Fuures Exchange (2002) on several of Ausralia s NEM jurisdicions. An examinaion of he relaionships beween Ausralian spo and derivaive elecriciy prices would hen be ineresing. References Ausralian Compeiion and Consumer Commission (ACCC), 2000. Infrasrucure Indusries: Energy. Commonwealh of Ausralia, Canberra. Bera, A.K., Higgins, M.L., 1993. ARCH models: Properies, esimaion and esing. J. Econ. Surveys. 7, 305 366.
12 A. Worhingon, A. Kay-Spraley, H. Higgs Bollerslev, T., Chou, R.Y., Kroner, K.F., 1992. ARCH modeling in finance: A review of he heory and empirical evidence. J. Econome. 52, 5 59. Brennan, D., Melanie, J., 1998. Marke power in he Ausralian power marke. Energy Econ. 20, 121 133. Brooks, C., Henry, O.T., Persand, G., 2002. The effecs of asymmeries on opimal hedge raios. J. Bus. 75, 333-352. De Vany, A.S., Walls, W.D., 1999a. Coinegraion analysis of spo elecriciy prices: Insighs on ransmission efficiency in he wesern US. Energy Econ. 21, 435 448. De Vany, A.S., Walls, W.D., 1999b. Price dynamics in a nework of decenralized power markes. J. Regulaory Econ. 15, 123 140. Dickson, A., Warr, S., 2000. Profile of he Ausralian Elecriciy Indusry. Ausralian Bureau of Agriculural and Resource Economics (ABARE) Research Repor No. 2000.7, Canberra. Dunne, P.G., 1999. Size and book-o marke facors in a mulivariae GARCH-in-mean asse pricing applicaion. In. Rev. Fin. Analysis. 8, 35 52. Engle, R.F., Kroner, K.F., 1995. Mulivariae simulaneous generalized ARCH. Econome. Theory. 11, 122 150. Huang, B.N., Yang, C.W., 2000. The impac of financial liberalizaion on sock price volailiy in emerging markes. J. Comp. Econ. 28, 321 339. Inernaional Energy Agency (IEA), 2001. Energy Policies of IEA Counries: Ausralia 2001 Review. Organisaion for Economic Cooperaion and Developmen, Paris. Joskow, P.L., Kahn, E., 2001. A quaniaive analysis of pricing behaviour in California s wholesale elecriciy marke during summer 2000. Energy J. 23, 1-35. Karolyi, G.A., 1995. A mulivariae GARCH model of inernaional ransmissions of sock reurns and volailiy: The case of he Unied Saes and Canada. J. Bus. Econ. Sa. 13, 11 25. Kearney, C., Paon, A.J., 2000. Mulivariae GARCH modeling of exchange rae volailiy ransmission in he European moneary sysem. Fin. Rev. 41, 29 48. Lucia, J.J., Schwarz, E.S., 2001. Elecriciy Prices and Power Derivaives: Evidence for he Nordic Power Exchange. Universiy of California Los Angeles Working Paper, Los Angeles. Naional Elecriciy Marke Managemen Company Limied (NEMMCO), 2001. An Inroducion o Ausralia s Naional Elecriciy Marke. NEMMCO, Melbourne. Naional Elecriciy Marke Managemen Company Limied (NEMMCO), 2002. WWW sie: <hp://www.nemmco.com.au/>. Accessed December 2002. Pagan, A., 1996. The economerics of financial markes. J. Fin. 3, 15 102. Robinson, T., 2000. Elecriciy pool series: A case sudy in non-linear ime series modeling. App. Econ. 32, 527 532. Robinson, T., Baniak, A., 2002. The volailiy of prices in he English and Welsh elecriciy pool. App. Econ. 34, 1487-1495. Tai, C.S., 2000. Time-varying marke, ineres rae and exchange rae risk premia in he US commercial bank sock reurns. J. Mul. Fin. Mgm. 10, 397-420. Tse, Y.K., Tsui, A.K.C., 2002. A mulivariae generalised auoregressive condiional heeroscedasiciy model wih ime-varying correlaions. J. Bus. Econ. Sa. 20, 351-362. Wilson, R., 2002. Archiecure of power markes, Econome. 70, 1299-1340.
Table 1 Summary saisics of spo prices in five Ausralian elecriciy markes NSW QLD SA SNO VIC Mean 33.0244 42.7055 57.9171 32.5624 35.5077 Median 26.4246 30.4117 38.9352 26.5121 25.3052 Maximum 388.2060 1175.5260 1152.5750 366.1698 1014.6010 Minimum 11.6533 13.2871 11.5225 11.0992 4.9785 Sd. Dev. 29.6043 60.8140 92.1549 27.8366 58.5227 CV 0.8964 1.4240 1.5912 0.8549 1.6482 Skewness 6.8871 11.6290 7.6208 6.8653 12.0381 Kurosis 66.2028 187.4572 69.3994 69.0835 179.8255 Jarque-Bera 127447 1052805 141362 138754 970003 JB probabiliy 0.0000 0.0000 0.0000 0.0000 0.0000 ADF es -5.5564-7.6672-8.8834-6.1225-8.2235 Noes: NSW New Souh Wales, QLD Queensland, SA Souh Ausralia, SNO Snowy Mounains Hydroelecric Scheme, VIC Vicoria. ADF Augmened Dickey-Fuller es saisics; CV of variaion; JB Jarque-Bera. Hypohesis for ADF es: H 0 : uni roo (non-saionary), H 1 : no uni roo (saionary). The lag orders in he ADF equaions are deermined by he significance of he for he lagged erms. Only inerceps are included. Criical values are -3.4420 a.01, -2.8659 a.05 and -2.5691 a he.10 levels. Figure 1 Daily spo elecriciy prices for five Ausralian markes, 13/12/1998 30/6/2001 400 1200 300 200 1000 800 600 100 0 7/01/99 1/17/00 8/04/00 2/20/01 NSW 400 200 0 7/01/99 1/17/00 8/04/00 2/20/01 QLD 1200 400 1000 800 600 300 200 400 200 0 7/01/99 1/17/00 8/04/00 2/20/01 SA 100 0 7/01/99 1/17/00 8/04/00 2/20/01 SNO 1200 1000 800 600 400 200 Noes: NSW New Souh Wales, QLD Queensland, SA Souh Ausralia, SNO Snowy Mounains Hydroelecric Scheme, VIC Vicoria. 0 7/01/99 1/17/00 8/04/00 2/20/01 VIC
Table 2 s for condiional mean price equaions NSW (i = 1) QLD (i = 2) SA (i = 3) SNO (i = 4) VIC (i = 5) CONS. ** 12.8966 6.8610 * 16.0313 11.3500 16.18667 18.8600 ** 12.2740 5.5630 11.2951 20.7400 a i1 0.0497 0.7556-0.0135 0.0951-0.0237 0.0844 0.5977 0.8215 0.0248 0.1749 a i2 0.0410 2.0470 *** 0.5118 0.1291-0.0658 0.2296 0.2046 2.2010 0.0321 0.4654 a i3-0.1159 5.5800-0.0529 0.3520 0.2493 0.1946 1.0097 5.6880-0.0344 0.6905 a i4-0.0548 0.2984-0.0131 0.0778-0.0265 0.0557 ** 0.7001 0.3884 0.0318 0.1425 a i5-0.1641 4.0450-0.0049 0.3352 0.0310 0.1113 0.4664 4.0390 0.3102 0.5095 Noes: NSW New Souh Wales, QLD Queensland, SA Souh Ausralia, SNO Snowy Mounains Hydroelecric Scheme, VIC Vicoria. Aserisks indicae significance a * - 0.10, ** - 0.05, *** - 0.01 level Table 3 s for variance covariance equaions NSW (j = 1) QLD (j = 2) SA (j = 3) SNO (j = 4) VIC (j = 5) b 1j *** 80.2657 16.6300 18.7260 59.5500 120.9672 124.3000 *** 71.3986 12.8500 75.8586 78.8900 b 2j 18.7260 59.5500 *** 336.6956 99.0900 41.1680 332.7000 17.1266 66.2000 31.8362 285.4000 b 3j 120.9672 124.3000 41.1680 332.7000 ** 635.0478 353.4000 * 120.0339 88.1800 229.8638 219.7000 b 4j *** 71.3986 12.8500 17.1266 66.2000 * 120.0339 88.1800 *** 67.6679 11.7500 ** 75.3265 41.9500 b 5j 75.8586 78.8900 31.8362 285.4000 229.8638 219.7000 ** 75.3265 41.9500 *** 295.1421 62.2100 c 1j *** 0.0985 0.0140 0.0997 0.1735 *** 0.0989 0.0278 *** 0.1013 0.0043 *** 0.0992 0.0221 c 2j 0.0997 0.1735 *** 0.1008 0.0198 0.1232 0.2944 0.0993 0.2777 0.0834 0.3979 c 3j *** 0.0989 0.0278 0.1232 0.2944 *** 0.0991 0.0216 *** 0.1021 0.0126 *** 0.0937 0.0211 c 4j *** 0.1013 0.0043 0.0993 0.2777 *** 0.1021 0.0126 *** 0.1046 0.0105 *** 0.0978 0.0175 c 5j *** 0.0992 0.0221 0.0834 0.3979 *** 0.0937 0.0211 *** 0.0978 0.0175 *** 0.0915 0.0249 g 1j *** 0.8047 0.0133 *** 0.8412 0.3192 *** 0.7839 0.0959 *** 0.8080 0.0001 *** 0.8034 0.0447 g 2j *** 0.8412 0.3192 *** 0.8051 0.0416 0.6520 1.3560 ** 0.8413 0.4615 0.8234 1.0580 g 3j *** 0.7839 0.0959 0.6520 1.3560 *** 0.8107 0.0309 *** 0.7868 0.0961 *** 0.8148 0.0263 g 4j *** 0.8080 0.0001 ** 0.8413 0.4615 *** 0.7868 0.0961 *** 0.8098 0.0128 *** 0.8056 0.0316 g 5j *** 0.8034 0.0447 0.8234 1.0580 *** 0.8148 0.0263 *** 0.8056 0.0316 *** 0.8119 0.0233 Noes: NSW New Souh Wales, QLD Queensland, SA Souh Ausralia, SNO Snowy Mounains Hydroelecric Scheme, VIC Vicoria. Aserisks indicae significance a * - 0.10, ** - 0.05, *** - 0.01 level Table 4 Ljung-Box ess for sandardized residuals NSW QLD SA SNO VIC Saisic 27.0100 32.4600 44.7000 17.9700 50.8700 p-value 0.0077 0.0012 0.0000 0.1166 0.0000