Cross-border auctions in Euroe: Auction rices versus rice differences Natalie Glück, Christian Redl, Franz Wirl Keywords Cross-border auctions, electricity market integration, electricity rice differences Abstract Basic economics suggest that rices for scarce interconnector caacity should reflect (if not equal) the arbitrage from trade, i.e., of low cost suliers selling to a high rice country. This aer analyses emirically whether and to what extent cross-border auction rices for interconnectors reflect the whole-sale electricity rice differences. The aer focuses on the following two interconnectors: The sea cable NorNed between the Netherlands and Norway, which came into oeration on May, 6th 2008 [1], connecting the North-Western-Euroean market with the NordPool market and the interconnectors between Germany (E.ON and RWE) and TenneT (the Netherlands). It turns out that the conjectured causality is weak and in fact the reverse causality (from interconnector rices to sot rice differences) seems more likely on statistical ground (Granger causality test). An exlanation is that the interconnector auctions recede the sot rice transaction and are thus an estimate for the arbitrage from trade. I. INTRODUCTION The Directive 2003/54/EC on common rules for the internal electricity market [2] and the Regulation EC/1228/2003 on cross-border exchanges [3] emhasise the utmost imortance of the exansion of transmission networks in order to create a common Euroean electricity market. However, the DG Cometition Reort on the energy sector [4] shows how insufficient this develoment has been so far and still many bottlenecks [5] exist such that wholesale electricity rices differ substantially across Euroe. The scarce caacity on the interconnectors has to be auctioned to comly with economic rinciles set in the Regulation. If the auction mechanism functioned and market actors have rational exectations, the caacity rice at the interconnector with scarce cross-border caacity should reflect the market rice difference since no market articiant is willing to ay more for the transfer caacity and since entry is free. Potential exlanations for lower rices are: transaction costs, risk aversion (not really convincing) and oligoolistic strategies that allow to influence and benefit from affecting the domestic rice, e.g., by retracting suly caacity which in turn makes exorts less attractive or even infeasible. However, a consistently remaining the arbitrage from trade is somehow a uzzle, since we (literally) could buy in one of the sot markets and resell across the border earning a rofit.
Alternative: Consider a rice taking sulier in market A, with the otions to sell at home at A and in country B at a rice B > A. Then clearly, the maximal willingness to ay for a right to exort one unit is equal to the rice difference, ( B - A ) if ositive, absent transaction, other transort costs, and risk aversion. Given cometitive bidding, the resulting auction rices should equal this maximal willingness to ay. Indeed, even under less ideal markets due to no-arbitrage condition this should be the case, since an auction rice below this difference would offer an arbitrage that anyone, even we, can exloit: Buying in the sot market and reselling at a rofit across the border. Therefore, the interconnector rice (IC) from country A to country B should deend on the differences in sot rices (if ositive, otherwise no flow in the direction from A to B makes sense and we know water does not flow uhill), (1) with α close to 1 if other costs (cross border fees, other transort charges to and from the border) are negligible. Some emirical work evaluating historical data until 2006 has shown, that a single market for electricity in continental Euroe had not been attained [6] and that the obtained rices of the exlicit transmission caacity auctions fail to exlain the international rice differences (see [6] for evidence from the continental Euroean market and [7] for assessments of the German-Danish cross-border auctions before market couling). Gebhardt and Höffler (2008) examine Euroean market integration by looking emirically and theoretically at German-Dutch and German-Danish interconnectors from 2000 to 2006 [8]. Including more recent data from 2008, this aer analyses by econometric means the hourly rices of the daily auctions of exlicit cross-border auctions at the following interconnectors: On the one hand the sea cable NorNed between the Netherlands and Norway, which came into oeration on May, 6th 2008 [1], connecting the North-Western-Euroean market with the Nordool market is analysed. On the other hand, the two connections between Germany and the Netherlands, namely the E.ON-TenneT and the RWE-TenneT interconnectors are addressed. Thereby the question, whether and to what extent the cross-border auction rices reflect the wholesale electricity rice differences between these three selected countries, shall be answered. Section II describes the data, followed by section III, with the exlanation of the three econometric models and the emirical methodology. In section IV the results of the regression analyses are shown. Finally, chater V concludes. II. THE DATA In this section the data sources and collected original data series of the sot market rices and interconnector rices are described and the adoted seasonal adjustments exlained. A. Sot Market Prices The Dutch electricity market is rimarily characterized by thermal generation, whereas Germany has besides high thermal generation also nuclear generation and large scale intermittent electricity roduction by wind. The Nordic market is dominated by run-of-river, umed-storage hydro and nuclear generation and has the lowest wholesale electricity rice as shown in Figure 1.
[ /MWh] For the analysis in this aer the day-ahead electricity market rices of the resective sot markets (EEX for Germany [9], APX for the Netherlands [10] and NordPool-Sot (the rice zone Kristiansand) for the Nordic market [11]) are taken for the year 2008. In a next ste the sot rice differences are calculated such that the hourly rices from the, in general, lower rice level country (A) are subtracted from the hourly rices of the, in general, higher rice country (B), as described in equation (2). SotBA t SotBA t B 0 A if if B B A A 0 0 (2) This transformation of the raw data accounts for the fact when an exort from A to B makes no sense B A since the home rice in A exceeds that abroad ( 0 ) which renders the urchase of an interconnector caacity in the direction from A to B unrofitable. 80 70 60 50 40 30 20 10 0 2000 2001 2002 2003 2004 2005 2006 2007 2008 Figure 1 Nordic DE NL Sot market base load rices [ /MWh] in Germany (DE), the Netherlands (NL) and the Nordic market from 2000 to 2008. Source: [9]-[11]. Figure 2 exemlarily shows the calculated hourly sot rice differences between the Netherlands and Norway. For all sot rice differences in the resective countries, only the ositive values are taken. This means that it is only looked at the direction where most of the trade has taken lace. As stated above, the NorNed cable came into oeration on May 6 th, 2008. In Figure 2 it can be seen that in the first months of oeration the sot rice differences were fluctuating more than later in the year. To revent misleading results due to the variable data in the beginning of the trading, the analysis is conducted with hourly data only from July to December.
1 87 173 259 345 431 517 603 689 775 861 947 1033 1119 1205 1291 1377 1463 1549 1635 1721 1807 1893 1979 2065 2151 2237 2323 2409 2495 2581 2667 2753 2839 2925 3011 3097 3183 [ /MWh] 250 Hourly sot rice differences between the Netherlands and the Nordool (rice zone Kristiansand, 06.05.2008-31.12.2008, Tuesday-Friday, excl. holidays) 200 150 100 50 0 Figure 2 Hourly sot rice differences between the APX Sot rice and the Nord Pool sot rice (rice zone Kristiansand), May December 2008, Tuesday-Friday, excl. holidays. Source: [10], [11] & own calculations. To eliminate the seasonality in the data series only the working days from Tuesday to Friday are taken into account (holidays which e.g. fall between Tuesday and Friday are also excluded) 1. Additionally, Unit Root Tests have been conducted for all data series to evaluate, if the data are non-stationary and therefore could cause surious regression results. However no unit root was found. B. Interconnector Prices and Available Caacity The interconnector caacities are auctioned by auction offices. For the NorNed sea cable, which came into oeration on May 6 th 2008, NorNed Auction [1] is the auction office jointly run by the two transmission system oerators (TSO) Statnett (Norway NO) and TenneT (NE). On the Dutch-German interconnectors, TSO Auction BV [12] has been conducting the auctions in both directions since sring 2000, also owned by the TSOs: TenneT (NE), RWE and E.ON (DE). 2 On each of the three interconnectors exlicit market auctions are executed. Hence, as illustrated in Figure 3 and Figure 4, which show the day-ahead auction time-lines for the NorNed auction and the Dutch-German auctions, sot market clearing for every hour of the following day takes lace several hours after cross-border auction closing time. 1 This is due to different rice atterns during the working days comared to weekends or holidays for both sot market and interconnector rices. 2 TSO Auction BV also runs the auctions between the Netherlands and Belgium, so also the Belgian TSO ELIA is taking art [12].
NorNed auction time-line: day-ahead 09:15 09:45 10:15 11:00 12:00 IC NoN and SotNNo for 24h of the next day Available caacity ublished on www.norned-auction.org Auction Closure Time Results ublished on www.norned-auciton.org IC NoN APX Gate Closure SotNNo Nord Pool Gate Closure Caacity auction rocess Sot market trading Figure 3 Timeline of the auction rocess between Germany (DE) and the Netherlands (NE). Source: [10]-[11]. NE-DE auction time-line: day-ahead 08:30 09:00 09:30 11:00 12:00 IC DN and SotND for 24h of the next day Available caacity ublished on www.tso-auction.org Auction Closure Time Results ublished on www.norned-auciton.org IC DN Caacity auction rocess APX Gate Closure SotND Sot market trading EEX Gate Closure Figure 4 Time-line of the auction rocess between Norway (NO) and the Netherlands (NE). Source: [10]-[11]. For the econometric analysis the hourly auction rices at each interconnector in 2008 are taken. But only the direction from the (mostly) lower rice level country (A) to the higher rice level country (B) is considered within the scoe of this aer. In these directions most of the trade during the year takes lace and the interconnector rice arameter is denoted by IC AB t. As an additional indeendent variable in the below described analysis, the hourly caacity available for the day-ahead auction rocess is taken as a third data series and denoted as Ca AB t. To account for the seasonality (for the same reasons as described above regarding the sot rices) the hourly interconnector rices are only taken for all working days from Tuesday to Friday and holidays were excluded. Now there are three articular interconnector rice arameters, namely interconnector from NO to NE, and E ON IC. NE RWE NE t and t NO IC NE t for the NorNed IC from E.ON- or RWE-Germany to NE. 3 3 Because data for the NorNed interconnector was only taken from July to December 2008 the resective sot rice data series are adjusted to that samle size.
Table 1 and Table 2 show the descritive statistics for the sot rice differences (SotDiff) and the several interconnector rices (IC). Table 1 Descritive statistics for the Norwegian-Dutch interconnector, sot rice differences (only ositive values) & theoretical otential for arbitrage for the original data series (May December) and the adoted data series (July-December). Descritive Statistics NO-NE (only ositive values; 2008: Tue-Fri, excl. holidays) (May-Dec) (May-Dec) Theoretical Arbitrage (May-Dec) (July-Dec) (July-Dec) Mean 42.5 39.9 2.58 36.0 33.9 Median 38.5 36.5 1.56 33.6 31.5 Maximum 190.6 193.5 87.96 159.5 193.5 Minimum 0.0 0.0-43.25 0.0 0.0 Std. Dev. 29.7 28.9 8.94 24.5 25.1 Observations 2,902 2,902 2,902 2,136 2,136 Table 2 Descritive statistics for the German-Dutch interconnectors & sot rice differences (only ositive values) and resective theoretical otential for arbitrage for the data series January December. Descritive Statistics NE-DE (only ositive values; 2008: 01.01.-31.12, Tue-Fri, excl. holidays) IC RWE-NE t Theoret. Arbitrage E.ON-NE Theoret. Arbitrage RWE-NE Mean 9.4 3.9 6.0 5.5 3.5 Median 5.9 1.7 1.5 3.3 3.2 Maximum 145.1 86.4 300.0 120.3 122.7 Minimum 0.0 0.0 0.0-40.0-300.0 Std. Dev. 12.5 7.3 21.3 10.0 22.1 Observations 3,002 3,002 3,002 3,002 3,002 In a next ste, the theoretical otential for arbitrage between the markets is calculated by Sot BA t IC AB t (3). Table 1 and Table 2 reveal a ositive otential for arbitrage on average larger at the German border - but one that is insignificant given the large standard errors. In the case of the NorNed cable - as exemlarily demonstrated in Figure 2 for the sot rice differences in the beginning of its oeration higher fluctuations in both the sot rice differences as well as the interconnector rices can be observed. Therefore, the adated data series from May to December are taken for the analyses described further below. In the following grah (Figure 5) the develoment of the monthly variances of the calculated theoretical arbitrage otential (hourly values) is shown. The value of the variances decreases from May until Setember and then remains constant at around the value of 6. This develoment also reflects the observed fluctuations in the sot rice differences and the interconnector rices in the beginning of the
oeration time of the NorNed cable, as mentioned above. Table 3 describes the descritive statistics of the monthly theoretical arbitrage otential at the NorNed cable. 16.0 14.0 12.0 10.0 8.0 6.0 4.0 2.0 Monthly variances of theoretical arbitrage otential: Sot NE-NO - IC NO-NE (2008) 0.0 May June July Aug Set Oct Nov Dec Figure 5 Table 3 Monthly variances of the theoretical arbitrage otential the differences between the sot rice differences and the interconnector rices. Source: [1], [10], [11] & own calculations. Descritive Statistics of the monthly theoretical arbitrage otential (hourly values) between Norway and the Netherlands. Source: [1], [10], [11] & own calculations. Descritive Statistics of theoretical arbitrage otential (NO-NE, 2008: SotDiff - IC) May June July Aug Set Oct Nov Dec Mean 5.5 2.5 2.1 3.3 1.6 0.5 2.5 1.6 Median 2.6 1.7 1.6 1.4 1.4 0.6 1.9 1.7 Maximum 88.0 57.3 69.4 41.1 21.3 27.0 37.7 36.5 Minimum -41.3-27.5-43.3-8.2-14.3-22.5-34.4-23.3 Std. Dev. 14.8 8.6 10.3 7.4 5.5 6.4 6.1 6.5 III. THE MODELS AND METHODOLOGY As mentioned in the introduction, the interconnector rice should erfectly reflect the sot rice difference, if the auction mechanism works (and even if does not work, actually!) and market actors have rational exectations. An alternative oint of view is based on the timing: Since the interconnector auctions recede the sot rice auctions by several hours (Figure 3 and Figure 4), one could view these auction results as redictors of the sot rice auction (as future market rices should redict sot rices). If this is the case and the sot market is non-cometitive, oligoolistic suliers will use this information to make their suly decision. In order to decide this direction of causality emirically, Granger-causality tests (as exlained in detail further below) were erformed, which, however, did not give clear directions.
A. Model 1: Arbitrage determines the interconnector rice The market articiants know the current electricity market rices and the resective differences of the rice levels in each market. They have exectations about the rofit they can gain by selling electricity to a market with a higher rice level. IC AB Caacity auction SotBA Volume trading Figure 6 Exectations about trading gains Schematic illustration of Model 1. The trader values the interconnector rice according the ossible rofits from trade. Hence, they value the caacity rices at the interconnectors according to their estimates of the ossible gains from trade, as illustrated in Figure 6 and described by equation 4. SotBA IC AB t c t (4) B. Model 2: Interconnector auction result drives the sot rice difference In exlicit auction, as alied on all of the here examined interconnectors, the auctions of the interconnector caacities take lace several hours before the auctions of the traded electricity volumes. This imlies that the market articiants (not only generators themselves but also electricity traders) lace their transmission caacities auction bids before the real volumes are auctioned in a searate ste (see Figure 3 and Figure 4). This suggests the alternative causality: the auction rices are an estimate, something almost like a forward rice, on the sot rice differences. If this drives the realization, then the interconnector rices are ahead of the sot rice realizations and thereby cause econometrically, but not economically (since it is only an estimate not an economic mechanism for) the sot rice difference. IC AB Caacity auction SotBA Volume trading Figure 7 Suly decision after aying for caacity Schematic illustration of Model 2. The trader makes his suly decision uon the aid rice for the interconnector caacity. Every trader is not willing to ay more than the exected sot rice difference for the auctioned caacity (see Figure 7). Therefore, AB SotBA t c IC t (5).
C. Granger non-causality test Before the regression analysis is conducted, Granger non-causality tests are conducted for 2, 10 and 15 lags. The results are shown in Table 4 and Table 5. Table 4 Granger non-causality tests (NO-NE). Original data: May-December 2008 Testing H 0 Probability Lags Sot NE-NO t 0.000 2 Data: July-December 2008 0.000 2 0.420 10 0.000 10 0.000 15 0.000 15 Probability Lags Testing H 0 0.018 2 0.000 2 0.910 10 0.000 10 Granger non-causality test 0.930 15 0.000 15 The test results for Norway-Netherlands taking the data from May to December do not give a clear direction excet for 10 lags. For the examles with 2 or 15 lags both null hyotheses ( does not Granger-cause SotNE-NO t; SotNE-NO t does not Granger-cause ) can either be rejected (if Probability < 0.05) or acceted (if Probability > 0.05). By taking 10 lags, the hyothesis SotNE-NO t does not Granger-cause has cleary to be rejected, and so it can be concluded that a causal relationshi from the sot rice difference to the interconnector rice exists. The Granger non-causality test of the data from July to December gives even a clearer outut. By taking 10 or 15 lags the same results as for the broader data range can be reorted, but at a much higher robability rate of more than 90%. Table 5 shows the results for the German-Dutch connections. It gives clear hints towards direction in the case of the RWE-NE interconnector clearly acceting the H 0 SotNE-NO t does not Granger-cause IC NO- NE t with a robability > 0.05. At the E.ON-TenneT interconnector, only by taking 10 lags the same H 0 can be acceted.
Table 5 Granger non-causality tests (DE-NE). E.ON-NE RWE-NE IC RWE-NE t IC RWE-NE t IC RWE-NE t Granger non-causality test Testing H 0 Testing H 0 Probability Lags 0.000 2 0.000 2 0.065 10 0.000 10 0.013 15 0.000 15 Probability Lags IC RWE-NE t 0.679 2 0.000 2 IC RWE-NE t 0.313 10 0.017 10 IC RWE-NE t 0.301 15 0.001 15 A. Model 1 IV. REGRESSION ANALYSIS In model 1 a regression is run where the interconnector rice is regressed against a constant and the sot rice difference. Table 6 shows the results for the NorNed cable, Table 7 and Table 8 for the two interconnectors from Germany to the Netherlands. The t-statistics show the significance of all estimation coefficients (c and α). The NorNed estimation results in a sot rice difference coefficient value close to 1, (but still not equal to 1). In contrast, the sot rice coefficients for the German-Dutch border are significantly below 1 and relatively small (in both cases, E.ON and RWE); furthermore the fitting is far worse than for NorNed. This is uzzling, because given the fact that all negative values have been set 0 (as described in chater II) and assuming rational exectations and erfect foresight, the regression coefficient of the constant term and the sot rice difference should not be statistically different from 0 and 1, resectively. However, in the case of NO-NE, the negative constants coefficient amounting to -1.29 can be interreted as fixed costs (e.g. transaction costs) to be aid in addition to the interconnector rice. Taking also into account the R-square values of the sot rice difference with the highest value 0.91 (NO- NE), not much influence on the interconnector rice can be gathered from the estimation outut in all three examles. The R-square is articularly low for the E.ON-Dutch interconnector.
Table 6 Model 1 - Estimations of the interconnector rice in 2008 (NO-NE). Deendent Variable: Variable Coefficient t-statistic Prob. 0.98 150.78 0.00 c -1.29-4.56 0.00 R2 0.91 R2 adj. 0.91 Durbin-Watson 0.49 Observation 2,136 Table 7 Model 1 - Estimations of the interconnector rice in 2008 (E.ON-NE). Deendent Variable: Variable Coefficient t-statistic Prob. 0.34 39.78 0.00 c 0.7 5.16 0.00 R2 0.35 R2 adj. 0.35 Durbin-Watson 0.51 Observation 3,002 Table 8 Model 1 - Estimations of the interconnector rice in 2008 (RWE-NE). 4 Deendent Variable: IC RWE-NE t Variable Coefficient t-statistic Prob. 0.39 25.35 0.00 c 1.16 4.75 0.00 R2 0.75 R2 adj. 0.75 Durbin-Watson 0.43 Observation 3,002 The Durbin-Watson statistic, which should be around 2 to be ok, is very low, and signals high autocorrelation (the same is the case for the equation results in Model 2 below). In Table 9 the resulting robabilities for each equation of the models are listed. They do suort the high autocorrelation signal of the Durbin-Watson statistic. Thus in a next ste, estimating an error correction model (AR-model) and testing of the variables lags may be aroriate. 4 In the estimation at the RWE-NE interconnector 11 dummy variables for outliers of the RWE-NE interconnector rice have been integrated.
Table 9 Residual tests of the estimation equations: In the left column the resective deendent variables of the tested models are listed. Residual Tests Deendent variables Serial correlation ARCH-Test White Heteroskedasticity Model 1: 0.000 0.000 0.000 0.000 0.000 0.000 IC RWE-NE t 0.000 0.000 0.219 Model 2: Probabilities Probabilities 0.000 0.000 0.000 E.ON: 0.000 0.000 0.000 RWE: 0.000 0.000 0.000 The first model reveals that the sot rice difference is not a sufficient estimator of the interconnector rice. In the following section the other way round is tested. B. Model 2 Table 10 to Table 12 show the estimation results for the three interconnectors in 2008 and how these rices influence the later sot market rice realization. In two cases (the NO-NE and the E.ON-NE case) the interconnector rices coefficients are nearly 1. But the constants coefficients are far from 0 in every case. Table 10 Model 2 - Estimations of the sot rice difference in 2008 (NO-NE). Deendent Variable: Variable Coefficient t-statistic Prob. 0.93 150.78 0.00 c 4.29 16.42 0.00 R2 0.91 R2 adj. 0.91 Durbin-Watson 0.54 Observation 2,136
Table 11 Model 2 - Estimations of the sot rice difference in 2008 (E.ON-NE). Deendent Variable: Variable Coefficient t-statistic Prob. 1.01 39.78 0.00 c 5.47 26.05 0.00 R2 0.35 R2 adj. 0.35 Durbin-Watson 0.97 Observation 3,002 Table 12 Model 2 - Estimations of the sot rice difference in 2008 (RWE-NE). 5 Deendent Variable: Variable Coefficient t-statistic Prob. IC RWE-NE t 0.45 25.35 0.00 c 7.24 32.14 0.00 R2 0.18 R2 adj. 0.18 Durbin-Watson 0.79 Observation 3,002 Both the R-squared values are very low - with the excetion of the NorNed case. The Durbin-Watson statistic is very low in model 2 too for all regressions (comare also the residual tests robabilities in Table 9). V. CONCLUSION Three models are examined. In the first model it is assumed that the traders look for arbitrage by selling from the low rice to the high rice country. Hence, the exected sot rice difference acts as the driver for the caacity rice. Model 2 uses the reverse causality, the caacity rice as a redictor for the sot rice difference. Although this is less convincing on economic grounds, it is rooted in the timing since the caacity auctions takes lace before the volumes are traded. In the third model, model 1 is amlified by an additional arameter, namely the available caacity at the considered interconnectors. Theoretically, both the framework of model 1 and model 2, are lausible. Which variable is finally influencing the other one, deends on the market articiants first consideration for building their estimates for a ossible rofit in cross-border trade. But no matter which of the two models is used, in theory there has to be a reflection between the interconnector rices and the sot market rice differences due to the fact that no market articiant is willing to ay more for the caacity as they exect the sot rice difference, which equals the ossible arbitrage from trade, to be. 5 In the estimation at the RWE-NE interconnector 11 dummy variables for outliers of the RWE-NE interconnector rice have been integrated.
Therefore it is all the more remarkable that the estimates do not suort it: None of the models and none of all three interconnector examles bear evidence for a reflection of the rice differences to the interconnector rices or vice versa. Possible exlanations could be found in comaring the number of articiants in the caacity auctions vs. the articiants in the energy auctions. Due to the fact, that articiating in an auction rocess is very time consuming and huge human resources are necessary, it can be assumed that only big (or dominant) market layers can coe with that and they could distort rices. Furthermore, the differences between the generated volumes in the markets and the transmitted volumes could be analyzed. Differences between them could have an imact on resulting rices. The data for the available and obtained caacity volumes at the cross-border interconnectors exist and has also been integrated as an additional indeendent variable in the equation formula. However, since the results have not shown much difference to the analysis conducted within that aer, the authors are not yet sure if the interconnector volume auctioned is in fact indeendent from the interconnector caacity rice. A two-stage least squares -test is foreseen as a next ste. And finally, as concerns the Dutch-German border, the long-term auctions (monthly, yearly) ossibly influence the day-ahead rices. If the long-term rices do not match the short-term rices (at least on average), arbitrage between these auctions can be assumed as exlaining factor; this is clearly an area of interest for future research. Summarizing, what started out with a very strong conjecture ended u in the documentation of a uzzle namely that aarently substantial additional cost elements (e.g. transortation cost rior to the crossborder transmission, transaction costs) or substantial arbitrage exist, in articular between Germany and The Netherlands. This requires additional research. VI. ACKNOWLEDGEMENTS The authors are grateful for valuable discussions to and comments from Gerd Solem and Ivar Wangensteen. Furthermore, we thank three anonymous reviewers for their helful comments and suggestions. This research was artially financed by the Klima- und Energiefonds and carried out within the scoe of the Austrian research rogram Energie der Zukunft. VII. REFERENCES [1] NorNed Auction: www.norned-acution.org (24.02.2009) [2] Directive 2003/54/EC of the Euroean arliament and of the council of 26 June 2003 concerning common rules for the internal market in electricity; Official Journal of the Euroean Union L 167/37. [3] Regulation (EC) No 1228/2003 of the Euroean arliament and of the council of 26 June 2003 on conditions for access to the network for cross-border exchanges in electricity; Official Journal of the Euroean Union L 176/1. [4] Euroean Commission (2007): DG Cometition reort on energy sector inquiry, SEC(2006) 1724, 10th January, Brussels.
[5] R. Haas, C. Redl, J. Knaek, H. Auer (2008): A single Euroean electricity market a vision? Energy Economics Grou, Vienna University of Technology, Austria. [6] G. Zachmann (2008): Electricity wholesale market rices in Euroe: Convergence? Energy Economics 30, 1659-1671. [7] T. Kristiansen (2007): An assessment of the Danish-German cross-border auctions. Energy Policy 35, 3369-3382. [8] G. Gebhardt, F. Höffler (2008): How to Determine whether Regional Markets are Integrated? Theory and Evidence from Euroean Electricity Markets. Not yet ublished and kindly rovided directly by the authors. [9] Euroean Energy Exchange AG, Augustuslatz 9, 04109 Leizig, Deutschland / Germany; Data available at: www.eex.com [10] Kindly rovided by: APX Grou - APX Grou Data Services, APX Data and Reorts; www.ax.com [11] NordPool Sot AS; Data available at: www.nordoolsot.com [12] TSO Auction BV; Data available at: www.tso-auction.org