Price convergence in the European electricity market

Transcription

1 Price convergence in he European elecriciy marke Dr. E. Dijkgraaf Prof.dr. M.C.W. Janssen Erasmus Compeiion and Regulaion insiue Erasmus Universiy Roerdam Augus 7 7 Conac: Elber Dijkgraaf SEOR-ECRi Erasmus Universiy Roerdam Room H 7-25 P.O. Box DR Roerdam The Neherlands dijkgraaf@few.eur.nl Phone: (31)

2 1. Inroducion In is decision in 1 s insance on he proposed merger of Nuon and Essen, he NMa reieraed he posiion on he relevan geographical marke for producion and wholesale ha was earlier aken in is Visiedocumen. 1 In ha documen, based o a large exen on he analysis of Brale (6) 2, i is argued ha for non-peak hours here is evidence for he posiion ha he geographical marke is cerainly larger han he Neherlands, bu ha insead for peak hours, he marke is cerainly resriced o he Neherlands. Using daa of 5, Bralle (6) shows ha wholesale elecriciy prices during peak hours in Germany, Belgium and he Neherlands differed. Moreover, hey used a SSNIP es ype of analysis o argue ha in peak hours he marke is indeed confined o he Neherlands. The analysis of he Brale group is saic in he sense ha i does no consider he price pah over he years. Thus, i very well may be ha wholesale elecriciy prices during peak hours beween differen Norh-Wes European counries sill differed in 5, bu ha here is a (srong) endency over he years for prices o converge. If here is price convergence, hen he differen markes are moving owards (full) inegraion. A saic analysis does no show his price pah over ime. Thus, he mos imporan quesion is wheher price convergence can be observed, and if so, if he process of price convergence can be expeced o coninue in he near fuure. A dynamic sudy of price convergence is more relevan for he NMa decision on he proposed merger han a saic analysis as he merger will ake place in he fuure and a dynamic sudy can shed much more ligh on he issue of geographical marke definiion a he ime he merger will ake place han a saic sudy using daa on 5. Furhermore, we have more recen observaions a our disposal due o he availabiliy of daa over he years 6 and 7. Price convergence is likely o be observed as he las years have shown considerable changes in how differen counries are conneced wih each oher. Firs, naional exchanges sared rading in he las years in Belgium, France, Germany and he Neherlands. Given he ime needed o build up experience and he low volailiy of exchanges a heir sar, i could be expeced ha his increasingly affecs inernaional inegraion. Second, on 21 November 6 he power exchanges of he Neherlands, Belgium and France are coupled. This means ha no arbirage beween hese markes is necessary anymore as he inerconnecion capaciy is opimally used in he bidding process. The exchanges assume ha here is a single marke which means ha only capaciy resricions can resul in price divergence. Third, inerconnecion capaciy is increased by invesmens in new connecion capaciy and measures (e.g. phase shifers) o use he exising connecions more efficienly. This paper sudies price developmens over he period January 2 - July 7 in he Neherlands, France and Germany o deermine wheher indeed a process of price convergence exiss. To do his, we go back o he original raw daa on he per hour 1 See NMa (7), 1e fase beslui, numbers 81-93, Nederlandse mededingingsauoriei, Den Haag and NMa (6), Visiedocumen concenraies energiemarken, Nederlandse mededingingsauoriei, Den Haag. 2 Brale (6), Facors affecing geographic marke definiion and merger conrol for he Duch elecriciy secor, The Brale Group, London. 2

3 one-day-ahead wholesale marke prices in he Neherlands, France and Germany. 3 We firs fi a flexible rend paern o hese daa and invesigae he paern of he difference in hese rends. The picures obained show a clear endency for prices o converge beween all hree counries. We hen use more sophisicaed ess o argue ha indeed he process of marke inegraion is such ha i is exremely likely ha in 9 he marke for he producion and wholesale of elecriciy is larger han he Neherlands. In such a broader marke, he merging firms Nuon and Essen will no have a dominan posiion. There exiss some older lieraure suggesing a process of price convergence for elecriciy prices in Norh-Wes Europe. For insance, using daa up o 4, Zachmann (5) finds clear convergence beween Germany and he Neherlands for 12 ou of 24 hours. 4 However, here was no full convergence a he end of his period, since significan price differenials for peak hours were sill presen. Armsrong and Galli (5) find ha also for peak hours he differenial decreased beween 2 and 4. 5 They used daa for Germany, France, he Neherlands and Spain. Compared o hese papers, we use an up-o-dae daase and a differen mehodology, bu essenially reach he conclusion ha he process of price convergence has coninued unil he presen day. The paper proceeds as follows. Secion 2 presens he mehodology and daa. In secion 3 resuls are presened for non-peak hours and in secion 4 for peak hours. Secion 5 discusses he relaion beween he srucural changes ha one can observe in he marke (increases in inerconnecion capaciy and coupling of markes) and he price convergence we find. Secion 6 concludes. 2. Mehodology and daa If elecriciy producion plans of wo counries belong o he same marke, wholesale prices should (approximaely) be equal beween he wo counries. If prices are very close o each oher, hen a SSNIP es analysis is no needed anymore o conclude ha he geographical marke definiion should include a leas hese wo counries. The reverse does no hold rue. As explained in he inroducion, his paper sudies wheher here is a process of price convergence beween he Neherlands, France and Germany. For he Neherlands, France and Germany we dispose of he per hour wholesale prices for he one-day-ahead marke over he period of 1 January 2 unil 11 July 7. These prices are aken from he elecriciy exchanges APX, PWXT and EEX. 3 We sudy he one-day-ahead marke for hree reasons. Firs, if here is price convergence in he dayahead marke and raders know his, hey will no accep differen prices in he forward markes. Second, forward markes are mu8ch more volaile and herefore difficul o inerpre. Third, Nuon and Essen have used forward markes o show price convergence and our sudy is complemenary o wha hey did. 4 Zachman, G. (5), Convergence of elecriciy wholesale prices in Europe?, Deusches Insiu für Wirschafsforschung, Berlin. 5 Armsrong, M. en A. Galli (5), Are day-ahead prices for elecriciy converging in coninenal Europe? An exploraory daa approach. CERNA Working Paper, Ecole Naionale Supérieure des Mines de Paris. 3

4 To undersand movemens in elecriciy prices, i is imporan o recall ha prices depend on he ineracion beween demand and supply a any given momen in ime as sorage of elecriciy is no possible. This means ha prices can flucuae significanly during one day and he analysis has o be conduced per hour. We have prices for all individual peak (he 9 h ill he h hour for weekdays) and non-peak hours. 6 The NMa and all previously menioned lieraure concludes ha for non-peak hours he marke is broader han he Neherlands. In hese hours demand is low resuling in enough inerconnecion capaciy o creae an inernaional marke. The main discussion is hus abou wheher an idenical conclusion holds for peak hours. This is no obvious as demand is (much) larger and inerconnecion capaciy migh be oo low. Therefore, we concenrae our aenion on he analysis of peak hours, alhough we show ha our approach reproduces he conclusion ha here is indeed an inernaional marke for non-peak hours. We sar our analysis by esimaing a flexible rend equaion rough he available price daa o ge a firs impression of how he daa of differen counries compare wih each oher. If here is a process owards convergence, we should find ha he differences beween he rends diminish. If price convergence is reached, he differences should disappear a all. The flexible rend equaion we use is he following: elecriciy 2 9 Pi, h, = α 1 T + α 2 T α9 T + α1 + ε (1) where P is he wholesale price for hour h in counry i a day and is measured in euros per MWh. T is a ime rend ha increases wih 1 for each day of our observaions (T=1 for 1 January 2 and 18 for 11 July 7). As we know a priori ha prices flucuae considerably over he years, e.g. as a resul of changes in demand or supply, we include several rend variables o arrive a he rue rend paern ha is available in he daa. The polynomial equaion (of degree 9) we use, allows for a sufficien degree of flexibiliy (ha could no be obained if a polynomial degree of much lower dimension is used). Moreover, we have checked ha higher order polynomials do no creae a significanly differen picure. We plo he esimaed price rends per counry and plo also he differences beween he prices in he Neherlands compared o he prices in France and Germany. Wih perfec price convergence hese differences should be zero. Posiive values reflec higher prices in he Neherlands compared wih Germany or France. The plos provide a clear firs impression of wheher a process of price convergence is presen. However, only a formal es makes i possible o arrive a a saisically robus conclusion wheher price convergence is presen. To his end, we use he following approach. We go back o he raw daa, ake he price raio beween he Neherlands on one hand and Germany and France on he oher hand and invesigae wheher here is a clear and significan (downward) rend available in he ime series. 7 We esimae 6 Our definiion of peak hours is based on he observed differences in prices beween Germany, France and he Neherlands and no on formal definiions of peak-hours. As we will observe laer in more deail, here is a smooh ransiion from non-peak o peak hours and vice versa, in he sense ha he price paern for hours 8 and 9 as well as he hours and 21 are very close o each oher. 7 A more formal way o es for price convergence is he Kalman filer approach. We applied his approach as a sensiiviy analysis and arrived a comparable conclusions as he ones presened in Secion 4. The Kalman filer approach is commonly used o sudy price convergence. See for an overview Grewal, M.S. and P. A. Andrews (1993), Kalman Filering Theory and Pracice, Upper Saddle River, NJ USA, Prenice Hall. 4

5 elecriciy elecriciy PAPX, h, PAPX, h, = α 1 T + α2 + ε and = α3 T + α 4 + ε. (2) elecriciy elecriciy PEEX, h, PPWXT, h, For a given α 2, if α 1 < hen prices in he Neherlands become smaller relaive o Germany over ime. For a given α 4, if α 3 < hen prices in he Neherlands become smaller relaive o France over ime. If his is he case one can hen esimae a which momen T in ime he Neherlands and Germany belong o one and he same marke, namely when α1 T + α 2 = 1. For he Neherlands and France marke inegraion is reached a he momen T when α3 T + α 4 = 1. As a sensiiviy analysis, we es for seasonal effecs by esimaing a version of (2) including fixed-effecs for monhs and weeks. Noe ha each ime series is prone o enormous flucuaions. Regularly, one observes high prices when emperaures are exremely low or high. Anoher reason for high price levels is he occurrence of an unexpeced fall-ou of an imporan power saion. One such an inciden occurred on 22 May 7, for insance, where a large nuclear power saion in Belgium sopped working for several hours. As his inciden happened a he end of our daa period we have inroduced a dummy variable o ake accoun of price differences ha day Price APX/EEX a Hour 17 Temperaure Figure 1. Price APX divided by price EEX (lef axis) and emperaure (righ axis) 3. Resuls non-peak hours We firs presen he resuls of he esimaed ime rend for non-peak hours based on he specificaion described in equaion (1). As an example, Figure 2 shows he esimaed ime rends for hour 2 based on he specificaion described in equaion (1). One can clearly see ha he esimaed price rends in he Neherlands, France and Germany are almos idenical. The same conclusion can be reached for he oher nonpeak hours (see Appendix A). In he Inroducion we have menioned ha oher 8 Wihou such a dummy variable he general conclusion abou he las period would be highly skewed because of his inciden. 5

6 sudies also find ha prices for non-peak hours have already converged for some years. Thus, one can conclude ha an inernaional elecriciy wholesale marke exiss for non-peak hours and ha he geographical marke definiion for non-peak hours should be larger han he Neherlands. As he analysis of non-peak hours does no yield any new resuls, we do no analyse hese hours in more deail Figure 2 Time rend resuls non-peak hours for hour 2: prices in euro per MWh 4. The analysis of Peak hours We now presen he analysis for peak hours. Firs, we analyse figures based on he flexible rend analysis described in equaion (1). As an example, Figure 3 presens he esimaed ime rend for hour There are a few ineresing observaions o be made on he basis of hese picures: (i) A nearly all days in he sample period, he ime paern of he series for Germany and France is nearly he same. This implies ha as far as wholesale elecriciy prices are concerned hese wo counries seem o behave as an inegraed marke; (ii) A he beginning of he period we analyse, he Duch wholesale elecriciy marke is no a all inegraed wih Germany or France. Prices a he Duch marke are significanly higher and if we would analyse he resuls for a single year we would find on average significan differences for 4 and 5. The resuls by Zachmann (5) and Brale (6), who found also significan price differences for peak hours in hese years, are herefore no surprising. However, he picures also show ha here is a rend owards price convergence as he differences clearly diminish in recen years. This rend is no analysed in Brale (6), bu is in accordance wih Zachmann (5). The peak of he price difference in 6 is lower han in 5 (and ha was already lower han he peak in 4) and also he 9 As he oher hours show a paern comparable wih hour 18 we include he figures for oher hours in appendix B. 6

7 period for which he esimaed price rends differ shorens significanly. In 6 he peak was very shor indeed Figure 3 Time rend resuls peak hours for hour 18: prices in euro per MWh Figure 3 indicaes a process owards marke inegraion as price differences diminish over ime and he period of price peaks also shorens. To es his more formally, we use Equaion (2) o esimae wheher a (negaive) rend exiss in he price raio series. Table 1 presens he esimaed coefficiens for he rends, he esimaed price raio found a 11 July 7 and also he prediced ime period a which marke inegraion will be reached (using saisical exrapolaion only). 1 Table 1 Esimaion resuls rend analysis of price raios Price raio APX versus EEX Price raio APX versus PWXT Hour Coefficien rend 1,2 Price raio a 11 July 7 Price convergence in: Coefficien rend 1,2 Price raio a 11 July 7 Price convergence in: *** 1.1 Aug *** 1.14 Mar *** 1.2 Aug *** 1.1 Feb *** 1.12 May *** 1.18 Aug *** 1.19 Oc *** 1. Jun *** 1.3 Oc *** 1.5 Nov *** 1.3 Aug *** 1.9 Dec *** 1.5 Nov *** 1.9 Feb ***.99 Jun *** 1.2 Aug *** 1.6 Jan *** 1.11 Jul *** 1.3 Jun *** 1.33 Jun ** 1.14 Apr *** 1.9 Mar * 1.9 Feb Aug 19 Noes: (1) Coefficien muliplied by 1, (2) *** / ** / * means significance a 1/5/1%. 1 We exclude observaions for weekends and observaions wih price levels lower han 1 euro per MWh (.3% of oal observaions) as hese observaions can resul in exreme price raios. 7

8 All rend coefficiens are negaive indicaing ha here is indeed a process of price convergence going on. All bu one esimaed coefficiens are significan a he usual confidence levels of 1% or 5%. The only hour where no significan effecs are found is in hour, which is a he border of he peak/non-peak hour disincion (see below). On he basis of he esimaed rends we can deermine he esimaed price raio beween he Neherlands and Germany and he Neherlands and France a he las day of our sample, July 11, 7. This esimaed price raio cleans he real price raio for emporary effecs. The raios range from.99 for hour 16 in he comparison beween he Neherlands and Germany ill 1.33 for hour 18 in he comparison beween he Neherlands and France. When exrapolaing he ime rends for he Neherlands and Germany we find ha marke inegraion will occur before early 9 for en ou of welve peak hours. Only for he hours 19 and i akes more ime for prices o converge. Comparing he prices for he 19 h and especially he h hour wih oher peak-hours shows, however, ha for hese wo hours he price differences are already quie small from he beginning of our sample period onwards. This means ha price convergence is no easily found for hese hours as rends are raher weak by definiion. In fac he analysis shows ha he price pah in hese hours reflecs more he price pah of nonpeak hours and for hese hours price had already converged by he beginning of our daa period. For he comparison wih France rapid price convergence is also found for mos hours so ha marke inegraion is expeced o occur as early as he firs monhs of 9 for en ou of welve peak hours.. Again, for hours close o he non-peak hours, in his case hours 9 and, prices in he Neherlands and France have been close o each oher from he beginning of he sample period onwards so ha he ime rend is weaker for hese hours. Table 2 Trend coefficiens alernaive specificaions rend analysis price raios 1,2 Price raio APX versus EEX Price raio APX versus PWXT Monh fixed effecs Day fixed effecs Monh fixed effecs Day fixed effecs *** -.17 *** -.11 ** -.11 *** *** -.5 *** -.46 *** -.46 *** *** -.43 *** -.46 *** -.46 *** *** -.4 *** -.62 *** -.63 *** *** -.35 *** -.39 *** -.39 *** *** -.56 *** -.54 *** -.54 *** *** -.42 *** -.41 *** -.41 *** *** -.42 *** -.42 *** -.42 *** *** -.43 *** -.3 *** -.3 *** *** -.45 *** -.52 *** -.5 *** *** -.11 *** -.16 *** -.15 *** -.6 ** -.6 ** Noes: (1) Coefficien muliplied by 1, (2) *** / ** / * means significance a 1/5/1%. To es for specificaion errors, we conduced a sensiiviy analysis o check for seasonal effecs. Firs, monh dummies are included o es for monh specific fixed effecs. These effecs are significan, especially for Augus, Ocober and November. The general rends remain significan and negaive. Their sizes are barely influenced by alernaive specificaions and are in mos cases a lile larger (compare Table 1 and 2). This means ha Table 1 on average underesimaes he speed of convergence 8

9 compared o esimaions where monh dummies are included. Second, week dummies are included o es for week specific fixed effecs. A wo decimals we find nearly no differences wih he specificaion wih monh dummies. 5. Srucural causes So far, we have only looked a he price daa wihou having any srucural view on wha may cause price convergence. Such a pure daa analysis should ideally be accompanied by a more srucural analysis uncovering he causal mechanisms underlying he price convergence resuls ha we found so far. Elecriciy markes in Norh-Wes Europe have undergone several srucural changes in he las few years (such as enlarged inerconnecion capaciy and a coupling of he naional power exchanges; see also he Inroducion). As inerconnecion capaciy will be increased and marke coupling of exchanges will occur again in he fuure, i is worhwhile o analyse he effecs of hese srucural changes. This is wha we will do in his secion and we mainly focus on he effec of inerconnecion capaciy and marke coupling of exchanges. Over he pas years, he inerconnecion capaciy beween differen counries has been enlarged for he one-day ahead exchange, making capaciy resricion a less urgen problem. Moreover, several measures were aken o use he exising inerconnecion capaciy more efficienly. Table 3 gives he available capaciy beween Germany and he Neherlands for he one-day ahead exchange. The available capaciy for he oneday ahead exchange has risen clearly during he years On average nearly 7% more capaciy was available in 6 compared o 2 during peak hours. However, in 7 he capaciy decreased again as Tenne allocaed more capaciy o he monh and year aucions a he expense of capaciy available for he one-dayahead marke. Table 3 Capaciy inerconnecion Germany Neherlands in MW Capaciy in MW for Eon Tenne plus RWE Tenne Hour This does no necessiae ha oal inerconnecion capaciy has increased. Even wih equal oal capaciy he capaciy for he one-day-ahead exchange can increase if less capaciy is used by buyers on he year and monh exchange. 9

10 To analyse he effec of inerconnecion capaciy on price convergence in a simple way, we esimae he relaionships: elecriciy elecriciy PAPX, h, PAPX, h, = α 1 I h, + α 2 + ε and = α 3 I h, + α 4 + ε, (3) elecriciy elecriciy P P EEX, h, PWXT, h, where I represens he inerconnecion capaciy beween Germany and he Neherlands a day for hour h. This relaionship is, of course, oo simple o capure all aspecs of he complex impac of inerconnecion capaciy on wholesale price differences. In realiy, when bidders a he aucion for inerconnecion capaciy do no expec ha capaciy is fully used, an increase in capaciy will no have any effec on price differences beween counries. When esimaing he above relaionship using he whole sample, we herefore underesimae he real impac of enlarging he inerconnecion capaciy on peak hours when his capaciy is fully used. Therefore, we also esimae he same equaion using only daa where he inerconnecion capaciy is really binding, which we ake o be when he price paid for inerconnecion is a leas equal o 1 per MW. 12 This is, on average, he case for % of our observaions for peak hours. Table 4 presens he esimaed marginal effecs for an assumed 1 MW increase in inerconnecion capaciy, where he second column for each comparison is based on an analysis where only he daa poins are used where inerconnecion capaciy is really binding. 13 If α 1 < (α 3 <) hen an increase in inerconnecion capaciy leads o lower prices for he APX compared o he EEX (PWXT). This is he case for all hours as all coefficiens are negaive and significan. As may be expeced, he coefficiens for observaions where capaciy is binding are higher. This means ha increasing capaciy has a larger effec on price convergence for hour/day combinaions where currenly a lack of capaciy is presen. Table 4 Effec inerconnecion capaciy Germany Neherlands on price raios Price APX/EEX Price APX/PWXT Hour All observaions Price inerconnecion > 1 euro per MWh All observaions Price inerconnecion > 1 euro per MWh *** -.9 *** -.4 *** -.9 ** *** -.17 *** -.12 *** -.17 *** *** -.15 *** -.13 *** -.16 *** *** -.11 *** -.15 *** -.16 *** *** -.11 *** -.9 *** -.12 *** *** -.15 *** -.13 *** -.17 *** *** -.12 *** -.1 *** -.14 *** *** -.17 ** -.11 *** -.19 ** *** -.12 *** -.8 *** -.11 ** *** -.23 *** -.16 *** -.23 *** *** -.7 ** -.4 *** -.8 *** -.2 *** -.6 ** -.2 *** -.7 *** Average Noe: *** / ** / * means significance a 1/5/1%. 12 We use he price beween RWE and Tenne for each hour/day combinaion. 13 On average he effecs do no change significanly when we include a rend in he esimaions. 1

11 These figures can bes be undersood in comparison o he earlier analysis where we esimaed he size of he average price raio in each peak hour beween he Neherlands and Germany on one hand and he Neherlands and France on he oher hand (Table 1). Comparing hese esimaes wih he esimaes presened in Table 4 allows us o calculae how much addiional capaciy is sill needed, on average, before we have full marke inegraion. The resuls of his calculaion are presened in Table 5. The necessary capaciy increase o reach full marke inegraion depends of course on wheher coefficiens are aken from he full sample or from he sample where capaciy is binding. In he firs case 4 MW is needed before he Neherlands and Germany are inegraed in all hours, while an addiional 427 MW of inerconnecion capaciy is needed for marke inegraion of he Neherlands and France. This is, however, probably an overesimaion of he increase in inerconnecion capaciy ha is really needed for marke inegraion as he effec of an increase in inerconnecion capaciy is larger when we resric he analysis o observaions where capaciy is binding. Now, only 211 MW addiional inerconnecion capaciy is needed o be able o speak of a fully inegraed marke beween he Neherlands and Germany and only 14 MW o inegrae he Neherlands and France. Clearly, relaively lile addiional inerconnecion capaciy is needed o be able o speak of full marke inegraion. Table 5 Necessary increase in inerconnecion capaciy Germany Neherlands in MW Capaciy needed o inegrae APX/EEX Capaciy needed o inegrae APX/PWXT Hour All observaions Price inerconnecion > 1 euro per MWh All observaions Price inerconnecion > 1 euro per MWh Max The already planned exensions of he impor capaciy are already larger han he numbers we have esimaed o be necessary o be able o speak of marke inegraion. 14 We know, for example, ha he planned insallaion of phase shifers (dwarsregelrafo s) will increase he inerconnecion capaciy wih Germany wih abou 5 MW in he near fuure. This increase alone is enough o arrive a a capaciy ha is necessary for full marke inegraion. Furhermore, he Norned cable wih a capaciy of 7 MW will be available in 8 and phase shifers will increase he 14 Noe ha we could no analysis he effec of connecions wih oher counries as daa were no available for a sufficien long ime period (France) or capaciy was no increased in he period 2-7 (Belgium). This means ha i is no sure wheher an increase in inerconnecion capaciy wih oher counries has he same effec as an increase wih Germany. 11

12 inerconnecion capaciy wih Belgium wih 3 MW. 15 These increases will simulae marke inegraion, reducing he chance even furher ha significan prices differences may exis a any poin in ime beween he Neherlands and oher counries. We herefore conclude ha in he nex wo years, even during peak hours, he geographical marke will be larger han he Neherlands. To analyse he effec of marke coupling we include in equaion (3) a dummy variable for he observaions since November 21 6, he firs day he Belpex, PWXT and APX were coupled. We expec he coefficien for his dummy o be significan and negaive, suggesing ha he price raios of APX compared o EEX and PWXT are lower. While his is obvious for he price raio APX/PWXT, an effec for he raio APX/EEX could also be expeced if he elecriciy markes of he Neherlands and Germany are inegraed indeed. Nearly all coefficiens are indeed significan and negaive (see Table 6). 16 Only for he APX/EEX raio for hour and for he APX/PWXT for hour 16 no significan effec is found. Comparing he effecs wih he effecs of increases in inerconnecion capaciy shows ha he effecs of marke coupling are quie subsanial. The effec for he APX/EEX raio is on average equal o an increase in inerconnecion capaciy wih 317 MW, while he effec for he APX/PWXT raio equals 278 MW. For individual hours his figure is even higher wih a maximum for hour 19 in boh comparisons (469 MW for he APX/EEX and 53 MW for he APX/PWXT). This shows ha marke coupling is indeed an effecive way o inegrae markes. Table 6 Effec marke coupling on price raios Hour Price APX/EEX Price APX/PWXT * -.13 * *** -.34 ** *** -.3 ** ** -.32 ** * -.21 * * -.33 ** ** -.26 ** * * -.26 ** ** -.51 ** ** -.22 *** ** Average Noe: *** / ** / * means significance a 1/5/1%. Table 3 shows ha he available inerconnecion capaciy decreased in 7 for he one-day-ahead marke wih on average 316 MW. Our rend analysis makes clear, however, ha his did no resul in less price convergence in 7. Apparenly he effecs of marke coupling, sared jus before less capaciy became available, fully compensaed for he decrease in inerconnecion capaciy. 15 See NMa (6), e.g. aricle 152, and Allen & Overy (7), Toeliching aangeboden remedies in kader van beoogde Essen/Nuon fusie, May The resuls for he effecs of inerconnecion capaciy on he price raios are no significanly influenced by including a dummy for marke coupling. 12

13 The effecs of he marke coupling of APX, PWXT and Belpex are ineresing as he coming years oher markes will be coupled as well. A he end of 7 a coupling is expeced beween APX and Nord Pool when The Norned cable is finished. Coupling wih he EEX is expeced for he near fuure. Given he resuls of he coupled APX, PXWT and Belpex, hese new iniiaives will probably resul in a faser process owards marke inegraion. 6. Conclusions The conclusion of his analysis is crisp and clear. A his very momen a fully inegraed wholesale elecriciy marke beween he Neherlands and Germany/France does no exis ye. However, he process of price convergence is so srong ha marke inegraion can be expeced o ake place in approximaely he nex wo years. The pure daa analysis poins in his direcion, as well as he analysis relaing price difference o available impor capaciy and o marke coupling. Over he las years available capaciy has increased considerably and his process of increasing impor capaciy will coninue in he near fuure when new connecion capaciy will become available. The same holds rue for marke coupling as he APX, Belpex and PWXT are expeced o couple in he near fuure wih oher European exchanges. Therefore, he srucural reasons for price convergence in he pas, will show up in an even sronger way in he near fuure. The srucural analysis also poins a full marke inegraion in he very near fuure. 13

14 Appendix A. Esimaed rends for non-peak hours: prices in euro per MWh Hour 1 Hour Hour 3 Hour Hour 5 Hour

15 Hour 7 Hour Hour 21 Hour Hour 23 Hour

16 Appendix B. Esimaed rends for peak hours: prices in euro per MWh Hour 9 Hour Hour 11 Hour Hour 13 Hour

17 Hour 15 Hour Hour 17 Hour Hour 19 Hour