Do German renewable energy resources affect prices and mitigate market power in the French electricity market ?

ABSTRACT

Several empirical studies show that renewable energy sources such as wind and solar power, typically supplied at low marginal cost, can cause electricity market prices to fall. Recent theoretical research and simulations also highlight the link between the integration of renewable energy and market performance in an oligopolistic energy market. This article looks at these dynamics in the context of cross-border effects between two highly interconnected electricity markets, France and Germany. Using a rich panel dataset for hourly data from November 2009 to July 2015, I estimate the impact of German wind and solar power production on both prices and market power in the French wholesale market. The findings highlight the importance of coordinating energy policies via joint renewable energy support schemes among interconnected European electricity markets.

KEYWORDS:

• Market power
• electricity markets
• renewable energy
• electricity prices

JEL CLASSIFICATION:

• C33
• D21
• D43
• L13
• L94

Disclosure statement

No potential conflict of interest was reported by the author.

Supplementary material

The supplemental data for this article can be accessed here.

Notes

¹ The inclusion of the rotation variable $Z=P\ast X$ in the demand function is crucial for identifying the degree of market power. If we excluded $Z$ from the demand function, Equation (7) would be rewritten as $P=\varphi Q+{\beta }_{w}W+\eta$ where $\varphi =({\beta }_q-\frac{\lambda }{{\alpha }_p})$. The supply relation is still identified but the degree of market power $\lambda$ is not. The $\varphi$ that we estimate cannot tell us the degree of market power $\lambda$ because it depends on both ${\beta }_q$ and $\lambda$, thus the supply relation (tracing market power) will be indistinguishable from the supply curve (representing perfect competition).

² The turnover in the French wholesale electricity market is used to estimate the demand function (8) (Hjalmarsson (2000); Bask, Lundgren, and Rudholm (2011)) despite its small proportion in the total load because it is subject to price elasticity unlike total load, of which a large proportion is sold at regulated tariffs in the retail market or is subject to long term transactions. However, in estimating marginal cost and the supply relation, it is vital to take full account of the total demand.

³ Other usages that consume electricity are industrial activities and transportation. However, aggregate industrial production is relatively stable for hourly and daily patterns. Seasonal variations are more apparent. This is largely controlled for in the model through the inclusion of dummy variables for season. Electricity consumption for transportation can be varied at hourly frequency but seems to exhibit less stochastic properties than other factors such as temperature. A large part of its variation is captured through fixed hours effects.

⁴ The carbon price represents an additional cost for electricity generated from fossil fuels. It may be either a direct cost, if CO2 allowances are purchased, or an opportunity cost, if allowances are received free of charge (De Perthuis and Jouvet (2011)). Thus, electricity producers add the carbon price to their marginal costs.

⁵ Even though the Fisher-type tests cannot reject the null hypothesis (presence of unit-root) of the gas price variable, the Levin-Lin-Chu and Im-Pesaran-Shin tests suggest a stationary variable.

⁶ Nickell (1981) shows that the dynamic panel models with fixed effects are biased by $(1/T)$ but as $T\to \mathrm{\infty }$, the fixed-effects estimator becomes consistent. Furthermore, the fixed-effects models seem to be a more appropriate specification for our dataset in which the individual dimension (hours) is relatively small because it would not lead to a loss of degrees of freedom. The results of the Hausman test gave the overall statistics, ${\chi }^2(7)$ for the demand `tion and ${\chi }^2(13)$ for the supply relation, having $p-value=0.000$. This led to the clear rejection of the null hypothesis that random-effects model provides consistent estimates.

⁷ This can be an issue when there is a one-day public holiday, which we would not expect to impact on demand the following day. That is why I introduced the variable Holiday which accounts for weekend and public holiday effects.

⁸ Note that the estimates given in Table 5 are weighted values; i.e. the average effects of wind and solar power do not distinguish two possible cases: when high solar production coincides with high prices during peak-hours, the cross-border effect is assumed to be substantially higher than in the case of low solar output during off-peak hours.

⁹ The results of robustness tests can be provided in supplementary materials upon requested.