Modelling Germany´s Energy Transition and its Potential Effect on European Electricity Spot Markets

The German Energiekonzept (Energy Concept) was proposed in 2010 with the goal of making the country one of the world’s most energy efficient and environmentally friendly economies (Bundesregierung, 2011). One year later, as a reaction to the multiple reactor meltdowns in Fukushima, this strategy was reinforced with a broad consensus within the German government to implement its Atomaustiegsgestz (Nuclear Phase-Out Act), by closing immediately eight nuclear power plants and then the remaining nine until 2022 (Bundesregierung, 2011). Subsequently, the Renewable Energy Source Act 2012 (RESA, 2012) aims to increase electricity generated from renewable sources to at least 35% by 2020 and to at least 80% by the year 2050. The RESA 2012 reaffirmed the basic principles of the feed-in tariff policy, which prioritizes renewable energy sources, the pledge to connect all renewable producers to the grid and the guarantee of a favourable unit price. This paper examines the potential impact of wind generated electricity produced in Germany on other European electricity markets, by employing MGARCH (multivariate generalized autoregressive conditional heteroscedasticity) models with constant and time-varying correlations for daily data. The interrelationship of electricity spot prices of APX-ENDEX (UK and Netherlands), Belpex (Belgium), EPEX (Germany and Switzerland), OMEL (Spain and Portugal), Nord Pool (Finland, Denmark and Norway) and Powernext (France) with wind penetration induced by the German system is studied from November 2009 to October 2012, thus covering the period before and after the closures of eight nuclear power plants. Literature Studies, such as Gross et al. (2006), Holttinen at al. (2009) and Smith et al. (2007) have highlighted the challenges associated with increased penetration levels of renewable energy sources as planned by the German government. There is, for example, a significant risk that a system with high wind power capacity will suffer electricity shortages and even blackouts. Other studies have shown that electricity spot market prices decrease to varying extents with the in-feed of wind generated electricity (see for example: Bode and Groscurt, 2006; Gil et al., 2012; Jacobsen and Zvingilaite, 2010; Neubarth et al. ,2006; Saenz de Miera et al., 2008; Sensfuß et al ,2008). The reduction of electricity spot prices is attributed to the cheaper wind generated electricity displacing offers from generators whose technologies have higher marginal costs (Sensfuß et al., 2008; Woo et al., 2011). Nevertheless, this positive effect may come at the cost of an overall increase in spot price volatility, due to the combined effect of non-storability of electricity and the high volatility of wind power (Woo et al., 2011; Milstein and Tishler, 2011; Green and Vasilakos, 2010). Despite integrated electricity markets being a promising instrument when managing intermittent sources of energy, the few studies that assess the volatility interrelationships among electricity markets have largely neglected the potential impact of renewables. Indeed, Bosco et al. (2007) remark that ‘[...] post-reform European price series have generally been studied in isolation and the issue of the interdependency in the price dynamics of neighbouring markets has largely been ignored.’ (p. 2). To date, a small body of literature applied a multivariate framework to electricity price volatilities (Worthington et al. (2005), Higgs (2009), Le Pen and Sévi (2010), Veka et al. (2012)). In this context, the present study aims to assess the potential effects of Germany´s energy transition on level and volatility of electricity spot prices in Germany and in other European countries. Germany serves as a statuary example, because of its increasing reliance on and investments in wind generated electricity as well as the size and importance of the German electricity market in Europe. Constant Conditional Correlation Bollerslev (1990) proposed a Constant Conditional Correlation MGARCH model (CCC), which has been preferred in empirical research over the BEKK specification because of its computational simplicity. This model is based on the decomposition of the conditional covariance matrix into conditional standard deviations and correlations. The conditional correlation matrix is time invariant and the conditional covariance matrix can be written for each time, t, as follows: H_t=D_t ΓD_t=ρ_ij (h_iit h_jjt )^(1/2) (1) Where 1≤i≤j≤K,t=1,…,N; K is the number of variables in the model and N is the number of observations in the estimation period; D_t=diag(h_11t^(1/2)…h_KKt^(1/2)), (2) Γ=ρ_ij (3) h_iit is the conditional variance of a univariate GARCH model and Γ is the symmetric positive definite constant conditional correlation matrix, with ρ_ii=1 ,∀i. Dynamic Conditional Correlation Although the CCC model overcomes the shortcomings of the BEKK and VEC models, the assumption of constant correlations may be too restrictive (Minović, 2009). Tse and Tsui (2002) and Engle (2002) therefore extended the CCC models to dynamic conditional correlation models (DCC), by including a time dependent conditional correlation matrix (Γ_t) and thus the conditional covariance matrix becomes: H_t=D_t Γ_t D_t (4) Where D_(t ) and h_iit are as defined in equation (2). Following, Tse and Tsui (2002) the conditional correlation matrix is given by: Γ_t=(1-θ_1-θ_2 )Γ+θ_2 Γ_(t-1)+θ_1 Ψ_(t-1) , (5) where 1≤i≤j≤K and θ_1 and θ_2 are non-negative constants such thatθ_1+θ_2<1 and. Γ, is the KxK symmetric positive definite constant parameter matrix with ρ_ii=1 for all i. Ψ_(t-1) is a function of the lagged standardized residuals ξ_it, and its ijth element can be denoted as: Ψ_(t-1,ji)=(∑_(m=1)^M▒ξ_(i,t-m) ξ_(j,t-m))/√((∑_(m=1)^M▒ξ_(i,t-m)^2 )(∑_(m=1)^M▒〖ξ_(j,t-m)^2)〗) where ξ_it=e_it/h_iit^(1/2) ` (6) Engle (2002) proposed the following alternative formulation: Γ_t=diag (q_11t^(-1/2)…q_KKt^(-1/2) )((1-θ_1-θ_2 ) Q ̅+θ_1 ξ_(t-1) ξ_(t-1)^'+θ_2 Q_(t-1) )diag(q_11t^(-1/2)…q_KKt^(-1/2) ), (7) where Q ̅ is the KxK unconditional correlation matrix of ξ_t, and θ_1 and θ_2 are non-negative parameters satisfying θ_1+θ_2<1 (Higgs 2009). The results of the MGARCH models indicate positive cross-market and lagged spillovers, as well as significant reduction in electricity spot prices with increasing wind penetration. Positive time-varying correlations between spot market volatilities are found for markets with substantial shared interconnector capacity, and wind penetration volatility is negatively associated with electricity spot price fluctuations. All in all, this study provides evidence that decisions made by one state in the European Union regarding its electricity sector can impact on neighbouring electricity markets.