Renewable Energy and the Pricing of Electricity Futures

Transcription

1 Renewable Energy and the Pricing of Electricity Futures Sebastian Schwenen (TU Munich) & Karsten Neuhoff (DIW Berlin) BELEC 2016, DIW Berlin 1 / 14

2 Motivation Much research on how renewable energy (wind, solar) impacts spot power prices (Woo et al., 2011; Graf and Wozabal, 2015) Limited research on how intermittent renewables affect futures prices We employ a simple hedging model to study price and cost risk with intermittent generation We derive and estimate the futures market equilibrium condition using EEX/EPEX data Need for understanding risk patterns and risk mitigation; relates to market design debate (e.g., Hogan, 2005) 2 / 14

3 Renewables change spot price profile Day-Ahead Price [in EUR/MWh] Wind and Solar Output [in GW] January 2012 January 2014 Fitted values Fitted values Figure: Hourly renewable output and spot prices in the German market. 3 / 14

4 Related research & existing findings Literature typically focusses on explaining the risk premium: futures minus expected spot price (p f p) Old world in Bessembinder and Lemmon (2002): Uncertainty introduced by random demand Longstaff and Wang (2004) confirm BL using PJM data We add cost shocks from intermittent generation to BL setting, albeit in simplified model Hedging in energy markets: Perez-Gonzalez and Yun (2013) show its value, Aïd et al. (2011) study vertical integration 4 / 14

5 Optimal forward positions for generation A generating firm m has profit: π m = p(q m f m ) c r q r c th (q m q r ) + p f f m (1) where p and q r are random variables Mean-variance preferences (Newbery and Stiglitz, 1981): E(U(π m )) E(π m ) A 2 Var(π m) with risk parameter A Optimal aggregate forward positions for generating firms: F m = D + (c th c r ) Cov( Q r, p) Var( p) + M pf p A Var( p) (2) Larger Cov( Q r, p) increases F m which decreases (p f p) 5 / 14

6 Hypotheses We examine two hypotheses: H1: The risk premium decreases as the covariance of renewable output and price increases H2: The risk premium increases with marginal costs of thermal generation Add controls for the spot price variance and spot price skewness (BL, not part of our model) 6 / 14

7 Data & empirical specification p f i,j p i }{{} Risk premium = β 0+β 1 Cov( Q r, p) i 1000 } {{ } H1: Negative + β 2c coal i + β 3c gas i + β 4Var( p) i + β 5Skw( p) i +β 6X +ω+ɛ i,j }{{}}{{}}{{} H2: Positive BL: Negative BL: Positive 7 / 14

8 Data & empirical specification p f i,j p i }{{} Risk premium = β 0+β 1 Cov( Q r, p) i 1000 } {{ } H1: Negative + β 2c coal i + β 3c gas i + β 4Var( p) i + β 5Skw( p) i +β 6X +ω+ɛ i,j }{{}}{{}}{{} H2: Positive BL: Negative BL: Positive Futures prices data from EEX: weekly and monthly peak contracts, 2012 until 2014 Plus data on delivery time: spot price, wind and solar output, load, MC of conventional Estimate different specifications; quarterly fixed effects, time trends, only for observations close to delivery... 7 / 14

9 Illustration of typical futures price observations Futures Price [in EUR/MWh] /1/2013 1/1/2014 3/1/2014 5/1/2014 Tradingday Figure: Futures prices for delivery in May / 14

10 Illustration of typical premia observations Risk Premium [in EUR/MWh] /1/2013 1/1/2014 3/1/2014 5/1/2014 Tradingday Figure: Risk premia for delivery in May / 14

11 A rough look at weekly futures Risk Premium [in EUR/MWh] Covariance(Renewable Output, Price) [in Thousands] Fitted values Figure: Week-ahead risk premia for delivery in 2014 peak-weeks. 10 / 14

12 Regression results for weekly peak futures Dependent variable: Risk premium (1) (2) (3) (4) (5) H1: Covariance( Q r, p) (0.008) (0.008) (0.009) (0.009) (0.009) H2: MC coal (0.036) (0.036) (0.042) (0.043) (0.043) H2: MC gas (0.026) (0.026) (0.030) (0.031) (0.030) BL: Price variance (0.002) (0.002) (0.002) (0.002) (0.002) BL: Price skewness (0.203) (0.203) (0.237) (0.250) (0.243) Weeks ahead Yes Squared No No No Quarterly FE No No No Yes No Time trend No No No No Yes Sample Full Full week-ahead week-ahead week-ahead N Standard errors in parentheses p < 0.05, p < 0.01, p < / 14

13 Regression results for weekly peak futures Dependent variable: Risk premium (1) (2) (3) (4) (5) H1: Covariance( Q r, p) (0.008) (0.008) (0.009) (0.009) (0.009) H2: MC coal (0.036) (0.036) (0.042) (0.043) (0.043) H2: MC gas (0.026) (0.026) (0.030) (0.031) (0.030) BL: Price variance (0.002) (0.002) (0.002) (0.002) (0.002) BL: Price skewness (0.203) (0.203) (0.237) (0.250) (0.243) Weeks ahead Yes Squared No No No Quarterly FE No No No Yes No Time trend No No No No Yes Sample Full Full week-ahead week-ahead week-ahead N Standard errors in parentheses p < 0.05, p < 0.01, p < / 14

14 Regression results for weekly peak futures Dependent variable: Risk premium (1) (2) (3) (4) (5) H1: Covariance( Q r, p) (0.008) (0.008) (0.009) (0.009) (0.009) H2: MC coal (0.036) (0.036) (0.042) (0.043) (0.043) H2: MC gas (0.026) (0.026) (0.030) (0.031) (0.030) BL: Price variance (0.002) (0.002) (0.002) (0.002) (0.002) BL: Price skewness (0.203) (0.203) (0.237) (0.250) (0.243) Weeks ahead Yes Squared No No No Quarterly FE No No No Yes No Time trend No No No No Yes Sample Full Full week-ahead week-ahead week-ahead N Standard errors in parentheses p < 0.05, p < 0.01, p < / 14

15 Regression results for monthly peak futures Dependent variable: Risk premium (1) (2) (3) (4) (5) H1: Covariance( Q r, p) (0.006) (0.006) (0.009) (0.009) (0.009) H2: MC coal (0.029) (0.029) (0.040) (0.036) (0.040) H2: MC gas (0.020) (0.020) (0.029) (0.027) (0.028) BL: Price variance (0.001) (0.001) (0.002) (0.001) (0.002) BL: Price skewness (0.153) (0.153) (0.220) (0.213) (0.227) Months ahead Yes Squared No No No Quarterly FE No No No Yes No Time trend No No No No Yes Sample Full Full month-ahead month-ahead month-ahead N Standard errors in parentheses p < 0.05, p < 0.01, p < / 14

16 Regression results for monthly peak futures Dependent variable: Risk premium (1) (2) (3) (4) (5) H1: Covariance( Q r, p) (0.006) (0.006) (0.009) (0.009) (0.009) H2: MC coal (0.029) (0.029) (0.040) (0.036) (0.040) H2: MC gas (0.020) (0.020) (0.029) (0.027) (0.028) BL: Price variance (0.001) (0.001) (0.002) (0.001) (0.002) BL: Price skewness (0.153) (0.153) (0.220) (0.213) (0.227) Months ahead Yes Squared No No No Quarterly FE No No No Yes No Time trend No No No No Yes Sample Full Full month-ahead month-ahead month-ahead N Standard errors in parentheses p < 0.05, p < 0.01, p < / 14

17 Regression results for monthly peak futures Dependent variable: Risk premium (1) (2) (3) (4) (5) H1: Covariance( Q r, p) (0.006) (0.006) (0.009) (0.009) (0.009) H2: MC coal (0.029) (0.029) (0.040) (0.036) (0.040) H2: MC gas (0.020) (0.020) (0.029) (0.027) (0.028) BL: Price variance (0.001) (0.001) (0.002) (0.001) (0.002) BL: Price skewness (0.153) (0.153) (0.220) (0.213) (0.227) Months ahead Yes Squared No No No Quarterly FE No No No Yes No Time trend No No No No Yes Sample Full Full month-ahead month-ahead month-ahead N Standard errors in parentheses p < 0.05, p < 0.01, p < / 14

18 Conclusion & discussion We model how renewable output risk impacts hedging needs and thus futures prices Premium increases for generation the more renewable output and price risk offset each other We support this finding using EEX/EPEX data Some weaknesses and scope for extensions: Is it reasonable to focus on representative market portfolio? What is the impact of renewable support schemes? Estimate spot prices to proxy expectations Room for more research on relatively young Energiewende futures: Phelix-Sun Peak Futures, Wind Futures 13 / 14

19 Thank you 14 / 14