Liquidity and Risk Premia in Electricity Futures Markets

Transcription

1 Liquidity and Risk Premia in Electricity Futures Markets IAEE Conference, Singapore, June 2017 Ivan Diaz-Rainey Associate Professor of Finance & Co-Director of the Otago Energy Research Centre (OERC) With Fergus Bevin-McCrimmon, & Greg Sise (Energy Link Limited)

2 Outline 1. Introduction 2. Literature Review 3. Methodology 4. Results 4.1. Liquidity 4.2. Risk Premia 5. Conclusions

3 1. Introduction NZ >50% all generation in the form of hydro NZ: electricity market (1996); textbook reform (Joskow 2006) locational marginal pricing or nodal pricing Mandatory pool (all trades must come through the market) EA concerns about large risk premia & illiquidity of hedge market Low depth of volume & Large spreads Policy interventions (suasion): Mandatory MM and maximum spread Literature on the liquidity of electricity markets Frestad (2009): Nord Pool futures liquidity Hagemann & Weber (2013): German intraday market Established lit on RP but liquidity as a risk factor? Finance lit Liquidity as a risk factor (priced) 3

4 Research questions: RQ and Contribution How has the liquidity of the NZ electricity futures market evolved over time, and have policy-induced interventions to increase liquidity succeeded?; What drives risk premia in the NZ electricity futures markets and is (il)liquidity a factor in these premia? Contributions Augment the literature on the determinants of risk premia incorporate measures of (il)liquidity First analysis of liquidity and risk premia in the NZ EM Have policy-induced interventions improved liquidity of the futures markets. NZ ETS 4

5 2. Literature Modelling Risk Premia Statistical risk factors Bessembinder & Lemmon (2002): Skewness (+ve) and variance (-ve) Bunn & Chen (2013): Mixed results due to under specification Physical market factors Douglas & Popova (2008): What causes skewness and variance to vary Physical market Botterud et al. (2010): Factors which influence hedging decisions Production costs: Bunn & Chenn (2013) & Fleten et al. (2015) Speculative investor behaviour Liquidity? liquidity risk affects asset prices (Acharya and Pedersen, 2005; Amihud and Mendelson, 1986b; Amihud and Mendelson, 1986b). 5

6 3. Methodology Liquidity Open Interest - Number of contracts outstanding in the market Trading volume simplicity in calculation and interpretation Amihud s Illiquidity best performer in Marshall et al. (2012) Illiquidity Ratio t = r t V t Determine the effectiveness of policy measures through structural break tests Bai and Perron (2003) Identify changes in structure using global information criteria 6

7 3. Methodology Risk Premia Examine the ex post risk premia: RP t = F t,t S T Determine drivers of premia using models of increasing complexity Statistical risk measures (Bessembinder and Lemmon, 2002) Physical market factors in a hydro generation context (Botterud et al., 2010) Underlying costs of production (Bunn & Chen, 2013) Speculative investor behaviour (Fleten, 2015) Liquidity 7

8 3. Methodology Modelling Premia Model 1: B&L RP t = α 0 + α 1 Skew t + α 2 Var t + α 3 S t + Model 2: Physical Model 3: Production and Speculation Model 4: Liquidity k i α 3+i RP t i + 3 i=1 α 3+k+i Q i + ε t RP t = α 0 + α 1 Skew t + α 2 Var t + α 3 S t + α 4 Demand t + α 5 Inflow t +α 6 Storage t + k i α 6+i RP t i + 3 i=1 α 6+k+i Q i + ε t RP t = α 0 + α 1 Skew t + α 2 Var t + α 3 S t + α 4 Demand t + α 5 Inflow t + α 6 Storage t + α 7 Carbon t + α 8 Oil t + α 9 Stock t + k 3 i α 9+i RP t i + i=1 RP t = α 0 + α 1 Skew t + α 2 Var t + α 3 S t + α 4 Demand t + α 5 Inflow t + α 6 Storage t + α 7 Liquidity + k i α 7+i RP t i + 3 i=1 α 9+k+i Q i + ε t α 7+k+i Q i + ε t 8

9 3. Methodology Variables Variable RP t Skew t Var t S t Demand t Inflow t Storage t Carbon t Stock t Oil t Liquidity t Description Difference between futures price and settlement spot price Spot price skewness over the past 7/30 days Spot price variance over the past 7/30 days Average spot price in the current quarter/over the past 30 days Deviations from historical average demand at time t Deviations from historical average hydrological inflows at time t Deviations from historical average hydrological storage at time t Log return of NZUs Log return of NZX50 Gross Index Log return of Dubai Fateh oil spot price Volume, Open Interest & Amihuds illiquidity ratio 9

10 3. Data Period: 2/10/ /12/2015 Sources: Energy Link Limited & Bloomberg Electricity market data & liquidity Futures trading commenced in July 2009 Future contracts traded relative to two nodes Benmore & Otahuhu Futures data volumes, OI & closing prices Three contract forms : Nearest quarter 1 year ahead 2 years ahead Physical market demand, hydrological inflows & storage Financial market data NZU, Dubai Fateh Oil & NZX50 log returns 10

11 Trading Volume Trading Volume Trading Volume 4. Liquidity: Volume Front End Volume Average Year Ahead Volume Average 0 10/2/ /2/ /2/ /2/ /2/ /2/ /2/ /2/ /2/ /2/ /2/ /2/ /2/ /2/ Years Ahead Volume Average /2/ /2/ /2/ /2/ /2/ /2/ /2/

12 Illiquidity Ratio Illiquidity Ratio Illiquidity Ratio 4. Liquidity: Illiquidity Ratio 0.25 Average Front End Illiquidity ratio 0.25 Average One Year Ahead Illiquidity Ratio Average Two Years Ahead Illiquidity Ratio

13 10/2/ /2/2009 2/2/2010 4/2/2010 6/2/2010 8/2/ /2/ /2/2010 2/2/2011 4/2/2011 6/2/2011 8/2/ /2/ /2/2011 2/2/2012 4/2/2012 6/2/2012 8/2/ /2/ /2/2012 2/2/2013 4/2/2013 6/2/2013 8/2/ /2/ /2/2013 2/2/2014 4/2/2014 6/2/2014 8/2/ /2/ /2/2014 2/2/2015 4/2/2015 6/2/2015 8/2/ /2/ /2/ Liquidity: Open Interest Front End 1 Year Ahead 2 Years Ahead NZ ETS 13

14 4. Liquidity Results Structural Breaks Expected: 5 th January 2010 EA introduces market making 3 rd October 2011 Reduction of maximum spread to 5% Results: Front End volumes increased in October 2011 and January 2012 Front End illiquidity decreased in October 2011 Other contract forms largely unaffected by change Implications: first policy did not work but second did Maturity Variable Positive Breaks Negative Breaks Variable Positive Breaks Negative Breaks Variable Positive Breaks Negative Breaks Front End Benmore Volume 17/10/2011 Illiquidity 4/10/2011 Open Int. 28/4/2011, 2/4/2012, 6/11/ /10/2014 Front End Otahuhu Volume 5/01/2012 Illiquidity 21/07/2011 Open Int. 5/1/2011, 2/4/2012, 10/1/2013, 27/1/ Year Benmore Volume 20/01/2012 Illiquidity 30/07/2012 Open Int. 19/5/2011, 25/4/2012, 2/4/ Year Otahuhu Volume 7/09/2012 Illiquidity 7/09/2012 Open Int. 23/3/2011, 28/2/2012, 1/7/2013, 18/7/ Years Benmore Volume 5/09/2012 Illiquidity 4/07/2012 Open Int. 1/4/2011, 2/4/2013 6/03/ Years Otahuhu Volume 5/09/2012 Illiquidity Open Int. 24/3/2011, 2/7/2012, 1/10/2013 5/01/

15 Realised Risk Premia Realised Risk Premia 4. Risk Premia Results Benmore Otahuhu Q1 Q2 Q3 Q Q1 Q2 Q3 Q Front End 1 Year Ahead 2 Years Ahead Front End 1 Year Ahead 2 Years Ahead Evidence of significant premia Time variant nature In general, premia greater for further out curves Premia consistently higher in Q2 and Q3 Lucia & Torro (2011): Dependent on the season which the contract matures Dry months; demand is highest and hydro storage/inflows lowest May be related to balance between generators and retailers relative hedging (Fleten et al., 2015) 15

16 4. Risk Premia Results B&L Model RP t = α 0 + α 1 Skew t + α 2 Var t + α 3 S t + k i α 3+i RP t i + 3 i=1 α 3+k+i Q i + ε t Contract Form Front End One Year Ahead Two Years Ahead Node Benmore Otahuhu Benmore Otahuhu Benmore Otahuhu Constant *** *** *** *** *** *** Skewness *** * Variance ** Spot Price *** *** *** ** ** Lagged RP *** *** *** *** *** *** Q ** *** *** *** *** Q ** *** *** *** *** Q ** *** *** R^ N Durbin-Watson Chi-square heteroscedasticity (pvalue) < < < < < < Notes: *** represents statistical significance at the 1% level, ** represents significance at the 5% level, * represents significance at the 10% level Front end uses average spot price over the quarter, others use average over past thirty days. Skewness and variance based upon past seven days for front end, past thirty days for others. Newey-West standard error corrections made to overcome heteroscedasticity and autocorrelation. Regressions run on daily data 16

17 4. Intertemporal correlation of prices Panel B: Correlations in Movements in Benmore Futures Prices qtr ahead 2 qtr ahead 3 qtr ahead 4 qtr ahead 5 qtr ahead 6 qtr ahead 7 qtr ahead 8 qtr ahead Average for all quarters Average for all quarters for Benmore spot Correlation of spot prices and futures price changes Spot price correlations dissipates beyond two quarters Futures price correlations large beyond this Market behaviour inefficient? 17

18 4. Risk Premia Results Physical Market RP t = α 0 + α 1 Skew t + α 2 Var t + α 3 S t + α 4 Demand t + α 5 Inflow t +α 6 Storage t + k i α 6+i RP t i + 3 i=1 α 6+k+i Q i + ε t Contract Form Front End One Year Ahead Two Years Ahead Node Benmore Otahuhu Benmore Otahuhu Benmore Otahuhu Constant *** *** *** *** *** *** Skewness ** Variance *** ** Spot Price *** *** *** ** Demand Deviations * Inflow Deviations *** *** Storage Deviations ** * * Lagged RP *** *** *** *** *** *** Q ** *** *** *** *** Q ** *** *** *** *** Q ** *** *** R^ N Durbin-Watson Chi-square < < < < < < heteroscedasticity (p-value) Notes: *** represents statistical significance at the 1% level, ** represents significance at the 5% level, * represents significance at the 10% level Front end uses average spot price over the quarter, others use average over past thirty days. Skewness and variance based upon past seven days for front end, past thirty days for others. Demand, inflow and storage deviations based upon the past week for front end, past four weeks for others Newey-West standard error corrections made to overcome heteroscedasticity and autocorrelation. Regressions run on weekly data 18

19 4. Risk Premia Results Production Factors RP t = α 0 + α 1 Skew t + α 2 Var t + α 3 S t + α 4 Demand t + α 5 Inflow t + α 6 Storage t + α 7 Oil + α 8 Carbon t + α 9 Stock t + i i=1 Contract Form Front End One Year Ahead Two Years Ahead Node Benmore Otahuhu Benmore Otahuhu Benmore Otahuhu Constant *** *** *** *** *** *** Skewness ** Variance *** ** Spot Price *** *** *** *** Demand Deviations Inflow Deviations *** *** Storage Deviations ** * * Oil * ** Carbon ** *** ** Stock Lagged RP *** *** *** *** *** *** Q ** *** *** *** *** Q ** *** *** *** *** Q * ** *** *** k α 9+i RP t i + 3 α 9+k+i Q i + ε t R^ N Durbin-Watson Chi-square heter.(p-value) < < < < < < Notes: *** represents statistical significance at the 1% level, ** represents significance at the 5% level, * represents significance at the 10% level Front end uses average spot price over the quarter, others use average over past thirty days. Skewness and variance based upon past seven days for front end, past thirty days for others. Demand, inflow and storage deviations based upon the past week for front end, past four weeks for others Newey-West standard error corrections made to overcome heteroscedasticity and autocorrelation. Regressions run on weekly data 19

20 4. Risk Premia Results Volume RP t = α 0 + α 1 Skew t + α 2 Var t + α 3 S t + α 4 Demand t + α 5 Inflow t + α 6 Storage t + α 7 Volume F t + k α 7+i RP t i + i i=1 Contract Form Front End One Year Ahead Two Years Ahead Node Benmore Otahuhu Benmore Otahuhu Benmore Otahuhu Constant *** *** *** *** *** *** Skewness ** Variance *** ** Spot Price *** *** *** *** Demand Inflows *** *** Storage ** * ** Volume *** Lagged RP *** *** *** *** *** *** Q ** *** *** *** *** Q ** *** *** *** *** Q * ** *** *** 3 α 7+k+i Q i + ε t R^ N Durbin-Watson Chi-square heteroscedasticity (p-value) < < < < < Notes *** represents statistical significance at the 1% level, ** represents significance at the 5% level, * represents significance at the 10% level Front end uses average spot price over the quarter, others use average over past thirty days. Skewness and variance based upon past seven days for front end, past thirty days for others. Demand, inflow and storage deviations based upon the past week for front end, past four weeks for others Newey-West standard error corrections made to overcome heteroscedasticity and autocorrelation. Regressions run on weekly data 20

21 4. Risk Premia Results Open Interest RP t = α 0 + α 1 Skew t + α 2 Var t + α 3 S t + α 4 Demand t + α 5 Inflow t + α 6 Storage t + α 7 OpernInterest F t + k i α 7+i RP t i + 3 i=1 α 7+k+i Q i + ε t Contract Form Front End One Year Ahead Two Years Ahead Node Benmore Otahuhu Benmore Otahuhu Benmore Otahuhu Constant *** *** ** *** *** *** Skewness * Variance *** ** Spot Price *** *** *** *** Demand Inflows *** *** Storage * * * Open Interest ** *** Lagged RP *** *** *** *** *** *** Q ** ** *** *** *** Q ** ** *** *** *** Q ** *** *** R^ N Durbin-Watson Chi-square heter. (pvalue) < < < < < < *** represents statistical significance at the 1% level, ** represents significance at the 5% level, * represents significance at the 10% level Front end uses average spot price over the quarter, others use average over past thirty days. Skewness and variance based upon past seven days for front end, past thirty days for others. Demand, inflow and storage deviations based upon the past week for front end, past four weeks for others Newey-West standard error corrections made to overcome heteroscedasticity and autocorrelation. Regressions run on weekly data 21

22 5. Conclusions & Policy implications Liquidity Descriptive: liquidity (illiquidity) increasing (decreasing) in all maturates Structural breaks and policy: mandatory market making did not improve liquidity but spread to 5% did but only front end Risk Premia Significant positive and negative, for all maturities. Premia consistently higher in Q2 and Q3 Physical market factors play a role in driving front end premia Spot prices significant power Suggest inefficient behaviour. Liquidity determinant of RP in some cases But only in the longer dates futures Important as may discourage LT hedging 22