Based on the theoretical model, we gain insights on price relations in sequential markets with differing product granularity and restricted market participation. In the context of the day-ahead and intraday auction, hourly products are traded si-multaneously with quarter-hourly contracts (τ ∈ 1, 2, 3, 4). The model suggests that the price formation in these markets may be illustrated according to Figure 4.4. The day-ahead supply curve reflects the aggregate marginal cost function (C0(q)) as mar-ket participation is considered to be unrestricted in the first marmar-ket. Additionally, the gradient of the intraday auction supply curve equals the gradient of the supply curve of unrestricted producers (au1). As we model intraday auction prices in terms of deviations from the corresponding day-ahead prices, we project the respective supply curve gradient on the day-ahead market clearing point as expressed by Equa-tion (4.12). Differences between the quarter-hourly and hourly mean of the residual demand (Dτ− D) are now transferred into movements along the 15-minute supply curve and directly yield quarter-hourly intraday auction prices.
When we transfer these relations to subsequent hours as depicted in Figure 4.5, one can observe a distinct pattern of prices. Prices for quarter-hourly products fluctu-ate around the respective prices for hourly contracts as illustrfluctu-ated by the green price time series. If the market participation in the intraday auction was not restricted, the gradients of the supply curves would be equal in both markets and prices would follow the curve of the fictitious quarter-hourly residual demand level as marked in blue.
Following the illustration in Figure 4.5, we observe three typical price movements:
6In more detail, it is below transaction costs.
82
4.2 Price Formation in the Day-Ahead and Intraday Auction
Q p
D
resD
1resD
2resD
3resD
4resp
Quarter-hourly prices Hourly price
Figure 4.4: Supply and demand in the hourly and quarter-hourly market
First, for an increasing residual demand, prices in the first quarter-hour are signifi-cantly lower compared to the respective prices in the last 15-minute time interval of the hour. Second, with a decreasing demand profile, we identify reverse relations.
Third, a flat demand profile leads to low price variation.
0 10 20 30 40 50 60 70
Residual demand [GW]
00:00 01:00 02:00 03:00 04:00 05:00 06:00 07:00
0 10 20 30 40 50 60
Price [EUR/MWh] unrestricted participation au1> a1
Figure 4.5: Exemplary residual demand profile and the resulting pattern for quarter-hourly product prices
So far, the model suggests that the high price volatility in sequential electricity
4 Price Volatility in Commodity Markets with Restricted Participation
markets is mainly driven by two aspects. First, quarter-hourly trade volumes are driven by sub-hourly deviations of the residual demand from the respective hourly means. Second, the high volatility of prices may stem from restricted participation.
As a result, the gradient of the quarter-hourly supply curve increases compared to the respective hourly one (au1> a1).
4.3 Empirical Analysis
We first seek to test the applicability of the theoretical model with respect to histor-ical data. We compare alternative specifications and conduct sensitivity analyses in order to support the hypothesis of robust and meaningful results. Second, we set up an empirical approach to quantify the impact of restricted participation on the price relations under analysis. Based on the respective results, we derive welfare implications. We choose an empirical approach rather than analyzing historical bid data due to two reasons. First, effects on the demand and supply side may be hard to aggregate when processing raw bid curve data. We target to account for both restricted participation on the supply side as well as a varying elasticity of wholesale demand (Knaut and Paulus, 2016). The approach adopted within this paper allows for estimating an aggregated coefficient. Second, the transformation of intraday auction bid data, which exhibits a pronounced stepped shape, into linear relations is a complex issue.
We analyze the time period from January 2015 until the end of April 2017. In the following, we first give a short overview on the respective data. We then describe our estimation approach and, finally, we depict and evaluate the empirical results.
4.3.1 Data
Due to the implementation of the intraday auction on December 9, 2014, the em-pirical analysis is based on data from January 2015 until the end of April 2017. A detailed list of all variables that are used is presented in Table 4.1. The table includes a brief explanation for each variable and the symbols which are used. Additionally, Table 4.2 provides information on the most relevant descriptive statistics.
Price data for German electricity markets was gathered from the European Power Exchange (EPEX SPOT SE, 2017b). In addition, we refer to the residual demand as a crucial expalantory variable which comprises two elements. First, we use forecasts
84
4.3 Empirical Analysis
Table 4.1: List of variables and references
Symbol Label Variable Measure Reference
pid at id auction price Uniform settlement price for a 15-minute product in the German intraday auction
EUR/MWh EPEX SPOT SE (2017b)
pd at day-ahead price Hourly German day-ahead auction price
∆Dtr es residual demand deviation
Dt; Dt load 15; load 60 Realization of the 15-minute load and the respective hourly mean
GW ENTSO-E (2017)
∆Dt load deviation Difference of the 15-minute load and the respective hourly mean
GW ENTSO-E (2017)
4 Price Volatility in Commodity Markets with Restricted Participation
for the renewable generation which are provided by the four German transmission system operators (TSOs) who are in charge of the reliable operation of the power system (EEX, 2017a). We refer to forecasted values as trades in the day-ahead and intraday auction take place one day before physical delivery and are therefore based on expectations with regard to the electricity generation from wind and solar power plants.
Table 4.2: Descriptive Statistics (Units according to Table 4.1, N refers to the number of observations)
Variable N Mean Std.Dev. Min 25% Median 75% Max
id auction price 81,663 31.49 15.89 -164.48 22.62 30.57 39.92 464.37 day-ahead price 81,663 31.42 14.12 -130.09 23.94 30.29 38.11 163.52 residual demand 15 81,663 41.67 11.09 0.95 34.36 41.59 49.58 73.00 residual demand 60 81,663 41.67 11.06 1.86 34.39 41.60 49.56 72.39 residual demand deviation 81,663 0.00 0.81 -12.27 -0.39 0.00 0.38 8.82
solar power 15 81,663 3.92 6.06 0.00 0.00 0.07 6.19 27.18
solar power deviation 81,663 0.00 0.51 -5.97 -0.04 0.00 0.03 4.48
wind power 15 81,663 9.58 7.47 0.30 3.83 7.43 13.23 39.56
wind power deviation 81,663 0.00 0.18 -1.60 -0.07 0.00 0.07 1.50
load 15 81,663 55.17 10.00 25.04 46.81 54.85 64.09 78.09
load deviation 81,663 0.00 0.76 -13.29 -0.35 0.00 0.35 9.39
Second, the residual demand depends on the electricity demand. We use data on the system load since load is commonly considered as the best proxy for electricity demand7. We use data on the realized8load which is published on the transparency platform of the European Network of Transmission System Operators for Electricity (ENTSO-E, 2017).
4.3.2 Empirical Estimations