Environnmental and Energy and in Europe: are there multiple bubbles?

Transcription

1 Environnmental and Energy and in Europe: are there multiple bubbles? Anna Creti 1 Marc Joëts 2 1 U. Paris Dauphine, LeDA-CGMP & Ecole Polytechnique 2 Banque de France FIME June 2018 (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June Banque 2018 de1 France) / 23

2 The context: the European Market for Pollution Permits The EU ETS is the first large scale CO2 emission trading system in the world. Created in 2003 by the EU Directive 2003/87/EC, the system entered into force in (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June Banque 2018 de2 France) / 23

3 The context: the European Market for Pollution Permits The revised EU ETS Directive (Directive (EU) 2018/410) entered into force on 8 April The overall number of emission allowances will decline at an annual rate of 2.2% from 2021 onwards, compared to 1.74% currently. The Market Stability Reserve (MSR) - to reduce the surplus of emission allowances will be substantially reinforced (Chaton, Creti, Sanin, 2018) Between 2019 and 2023, the amount of allowances put in the reserve will double to 24% of the allowances in circulation. The regular feeding rate of 12% will be restored as of From 2023 onwards the number of allowances held in the reserve will be limited to the auction volume of the previous year. Holdings above that amount will lose their validity. Other rules to limit leakage, foster low carbon innovation and energy sector modernization (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June Banque 2018 de3 France) / 23

4 The context: the European Market for Pollution Permits Price of the one month future contract: historical values High volatiliy, strong decreasing trend (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June Banque 2018 de4 France) / 23

5 The context: the European Market for Pollution Permits Price of the one month future contract: from Phase II until today higher prices, as actors anticipate tighter constraints in the future (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June Banque 2018 de5 France) / 23

6 Speculation: beyond carbon price drivers? We assume that permits are financial assets and apply a modified asset pricing relationship, as permits do not generate dividends and must be hold by each installation to grant annual compliance to the emission cap. Paul Krugman - New York Times Blog-July 21, 2009 Is the threat of speculation a reason to shun cap and trade? EU carbon edges higher, may correct on lack of utility support LONDON, May (Reuters) - European carbon rose for a third consecutive day on Thursday, but traders said the speculative buying that was driving prices higher was not supported by utilities, heightening the possibility of a correction. EU carbon falls on speculative selling, lack of utility buying LONDON, May (Reuters) - European carbon prices shed more than 3 percent on Wednesday on speculative selling and a dearth of utility buying, as the market prepared for government auctions to resume on Thursday after a day s hiatus. (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June Banque 2018 de6 France) / 23

7 Research Question and Main Results The novelty of the paper is the attempt to disentangle fundamental price drivers from speculative attacks, dating these latters We adapt recent test developed by Phillips, Wu and Yu (2011, hereinafter PWY), Phillips and Yu (2011, hereinafter PY), and Phillips, Shi and Yu (2013, hereinafter PSY) who propose a recursive right sided unit root approach to detect and date-stamp mildly explosive processes Our analyis unveils two main results: Bubbles exist in each of the three EU ETS phases. Bubbles are generally short and occur less often than 5 times a year. Running the tests on the full sample, we do not find evidence of bubbles in the period Eight out of the eleven bubbles found occur in Phase II (2009 and 2011). These episodes are not explained by the energy fundamentals (oil, coal, gas, S&P500), but by some uncertainty due to environmental and energy policy announcements (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June Banque 2018 de7 France) / 23

8 Outline of the presentation Related literature The model: data and empirical strategy Bubbles: when, how long, by how much Bubbles determinants Robustness checks (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June Banque 2018 de8 France) / 23

9 Related literature Three strands of literature Carbon Price Drivers and Energy Fundamentals: Aatola et al. 2013; Alberola et al. 2008, 2009; Benz & Trück 2009; Bredin & Muckley 2011; Chevallier 2009, 2011; Conrad et al. 2012; Creti et al. 2012; Daskalakis et al. 2009; Fezzi & Bunn 2009, Hintermann, 2017 etc... Carbon Price, Uncertainty, Speculation: Schennach 2000; Chesney and Taschini 2012; Colla et al. 2012; Chevallier et al., 2015; Palao Sánchez et al. 2016; Hintermann (2015) and André (2015) develop models in which CO 2 price manipulation is driven by the oligopolistic structure of the polluting sectors; Lucia et al. (2015). Financialisation of Energy Markets: Choi & Hammoudeh, 2010; Creti et al., 2013; El Haouri et al. 2013; Mensi et al., 2014; Sadorsky, 2014; Xiaoli et al., 2013; Zhang, 2018 for a review. (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June Banque 2018 de9 France) / 23

10 The model: data Our database regroups the daily settlement permit prices of the European Climate Exchange (ECX) of the Intercontinental Exchange Group (ICE). We use ICE futures prices, relying on the trading volumes, the continuity of the price series, and its dominant application in the literature (for example, Rickels et al 2015, Chevallier 2009 and 2011, Joyeux and Milunovich, 2010; Bredin and Muckley, 2011; Lucia et al. 2015; Sanin et al. 2015). December 2008 futures contract is used for the year (Phase I), the December 2013 contract for the years (Phase II) and the December 2020 contract for (Phase III). Our database spans from April 27th 2005 to February 5th (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June 2018 Banque de 10 France) / 23

11 The model: theoretical framework Modified asset price model: general asset pricing equation: P t = ( ) 1 i E t (D t+i + U t+i ) + B t, (1) i=0 1 + r f where P t, D t, r f are respectively the price of the asset (after-dividend), the dividend and the risk-free interest rate. U t and B t are the unobservable fundamental and the bubble components of the asset price. In the case of no bubbles (i.e. B t = 0), the asset dynamic is driven by the dividend and the unobservable fundamentals. (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June 2018 Banque de 11 France) / 23

12 Bubbles are defined by the following process: E t (B t+1 ) = (1 + r f ) B t, Given the specific nature of pollution permits (compliance), Equation (1) becomes the following simplified asset pricing relation: P t = F t + B t, (2) where F t is the fundamental variable of carbon price and B t bubbles at time t. Three components are included in the fundamental variable F t, namely energy prices, economic activity, and abatement opportunities. F t is expressed by the following equation which is the first principal component of carbon drivers. It allows us to explain more than 60% of the four considered drivers F t = 0.520gas t brent t switch t 0.260stock t. (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June 2018 Banque de 12 France) / 23

13 The model: empirical strategy (I) Bubbles and asset pricing Phillips, Wu and Yu (2011), Phillips and Yu (2011), and Phillips, Shi and Yu (2013): recursive right sided unit root approach to detect and date-stamp multiple mildly explosive processes improvement over the rational bubble literature (Flood and Garber, 1980; Flood and Hodrick, 1986; Diba and Grossman, 1987 and 1988; and Evans, 1991) as well as some alternative approaches (Hall, Psaradakis and Sola 1999, Bhargava, 1986,Kim, 2000, Busetti and Taylor, 2004) Null hypothesis: the carbon is a financial asset and its price P t follows a random walk with an asymptotically negligible drift P t = dt η + P t 1 + ε t (3) d =constant, T =sample size, η=coeffi cient controlling the size of the constant term and the drift, and ε t iid ( 0, σ 2 ). In the alternative case, asset price is governed by a mildly explosive root. (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June 2018 Banque de 13 France) / 23

14 The model: testing procedure (II) Testing procedure based on a rolling window regression: P t = α r1, r 2 + β r1, r 2 P t 1 + k i=1 P t = P t P t 1, ε t iid N ( 0, σ 2 ), k = lag order. γ i r 1,r 2 P t 1 + ε t (4) r1 th and r2 th = fractions of the sample; r 2 = r 1 + r w and r w = fractional window size of the regression. We use two tests: The Sup ADF (SADF) test proposed by PWY: r 1 is fixed to 0 and the end point r 2 varies from r 0 (the smallest sample window size) to 1: SADF (r 0 ) = sup ADF r 2 0 (5) r 2 ɛ[r 0,1] For forward recursive right-tailed ADF test, the Generalized Sup ADF (GSADF) test extending PWY s approach by adding more flexible windows: { GSADF (r 0 ) = sup ADF r } 2 r 1 (6) r 2 ɛ[r 0,1],r 1 ɛ[0,r 2 r 1 ] (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June 2018 Banque de 14 France) / 23

15 The model: data-stamping (III) We perform both the Backward sup ADF (BSADF) approach and evaluate sup ADF test on a backward expanding sample sequence (end point of each sample fixed at r 2 and the starting point varies from 0 to r 2 r 0 ) BSADF r2 (r 0 ) = sup r 1 ɛ[0,r 2 r 0 ] { ADF r 2 r 1 } based on repeated implementation of the ADF test for each ending point: SADF (r 0 ) = sup {ADF r2 } r 2 [r 0,1]...and the more flexible GSADF test over each ending point r 2 [r 0, 1] GSADF (r 0 ) = (7) (8) sup {BSADF r2 (r 0 )} (9) r 2 [r 0,1] The origination and termination dates of a bubble are calculated by the first and the last chronological observations whose ADF statistics exceed and go below the critical value. (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June 2018 Banque de 15 France) / 23

16 The model: small sample and heteroskedasticity (IV) We complete our empirical strategy by using: As in Xiaoli et al. (2013), the recursive wild bootstrap method of Gonçalves and Kilian (2004) to account for conditonal heteroskedasticity Tests are applied to the original daily price series but empirical distribution are simulated from wild bootstrap sample that mimics the conditional heteroskedasticity of the original data We run PWY and PSY on different price series, both considering separately each phase and the three phases altogether. (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June 2018 Banque de 16 France) / 23

17 Results(Ia) Bubbles over the entire sample Data: summary stats (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June 2018 Banque de 17 France) / 23

18 Results(Ib) Bubbles over the entire sample h=length of explosive episode for a minimum of 3, 5 and more than 5 days resp. Parentheses: size of the bubble from origination to peak (trough); brackets: misalignment of the price at the end of the bubble relative to the random walk. (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June 2018 Banque de 18 France) / 23

19 Results (II) Bubbles determinants-energy markets F t does not present mildly explosive processes (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June 2018 Banque de 19 France) / 23

20 Results (III) Bubbles determinants-institutional Uncertainty and Potential for Speculation For each bubble date, analysis of the institutional framework: energy and environmental policy announcements at the European level this issue is empirically analyzed for example by Benz & Trück, 2009; Mansanet & Sanin, 2015, Koch et al 2016, Ying 2017 (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June 2018 Banque de 20 France) / 23

21 Results (IV) Event study For each bubble date, we have conducgted event studies which result non significant Bubbles are temporary divergence from the fundamentals, not necessarily statistical breaks in prices or abnormal returns. They are sign of (perhaps mild) speculation. (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June 2018 Banque de 21 France) / 23

22 Robustness Checks Bubble testing and data-stamping by phases Test procedure with and without bootstrap, and with different drift values Different definitions of the composite fundamental variable More insights on the trading behavior and speculative attacks: banking behavior in 2009; utilities buying and selling strategies (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June 2018 Banque de 22 France) / 23

23 Concluding remarks New regulation of carbon markets? as a measure of last resort to avert or deflate speculative bubbles, the European Environmental regulator could be mandated to intervene directly in exchange trading on an occasional basis by buying or selling future contracts, instead of deciding quantitative measures for the allocation adjustment? tighter restrictions on financial participants, public disclosure of harmonised position information as light-hand regulation measures Thank you! (U. Paris Bubbles Dauphine, LeDA-CGMP & Ecole Polytechnique FIME June 2018 Banque de 23 France) / 23