The effect of variable renewable energy sources on electricity price volatility: the case of the Iberian market

ABSTRACT

The relationship between variable renewable energy supply (V-RES) and electricity price volatility is a controversial issue in the economic literature. In general, the literature has been inconclusive about the sign of the impact of installed capacity of these technologies on price volatility. This paper investigates the impact of V-RES on price volatility for the Iberian Market of Electricity (MIBEL), in the period ranging from 2010 to 2015. Using regression analysis and EGARCH models, we conclude that V-RES, and more specifically wind power supply, heightens price volatility. Likewise, greater intraday variability of V-RES also induces higher price volatility. Finally, following an analysis of the connection with the French market, we find that market coupling could help alleviate the sensitivity of price volatility to wind power supply variability.

KEYWORDS:

• Electricity
• wind power
• solar power
• price volatility
• renewables
• market coupling

JEL CLASSIFICATION:

• Q41
• Q42
• Q48

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Paulo Pereira da Silva http://orcid.org/0000-0003-4774-128X

Notes

1 V-RES have near-zero marginal costs, followed by nuclear energy, lignite, hard coal, gas and fuel oil plants (Cludius et al. 2014).

2 Authors argue that wind speeds are almost uniformly distributed throughout the day in Denmark; by contrast, in Germany, wind power output augments during off-peak hours, causing prices to plunge and heightening volatility.

3 Enrica De Cian, Elisa Lanzi, and Roberto Roson (2013) show that temperature and temperature variability shape the demand both in the long-run and in the short-run, along with other factors, such as income and energy prices.

4 In 2015, the share of RES (wind, solar PV, thermal, and others) in terms of electricity generation hovered around 29.2% and 31.1% for Spain and Portugal, respectively, while fossil energies had a share of 40.2% and 38.3% of electricity production. Spain achieved a global top position in terms of installed capacity of wind and PV power.

5 Indeed, there are several assessments focusing on the impact of market coupling on electricity markets. Notwithstanding, none of those studies delves into the effect of coupling on the association between volatility and V-RES output, which is the novelty of our study. For instance, Keppler, Phan, and Le Pen (2016) show that market coupling and the establishment of a combined order book on the basis of information of both markets contributed to the decline of the spread between French and German electricity prices. Schmid and Knopf (2015) show that pan-European transmission capacity expansion constitutes a no-regret option for integrating increasing shares of V-RES. Oggioni and Smeers (2013), Ochoa and van Ackere (2015), Pellini (2012), and Lam, Ilea, and Bovo (2018) also appraise the advantages and disadvantages of market coupling.

6 The load is equal to the electricity demand.

7 European regional market for Spain and Portugal. www.omie.es (last accessed April 2017).

8 Like Rintamäki, Siddiqui, and Salo (2017), demand and supply forecasts are utilised rather than realised values, because only forecasts are available for market participants when determining their bids to the day-ahead market.

9 In a non-tabulated analysis, we re-run our regressions introducing binary variables that assume the value of one when congestion is present and interaction variables of those dummy variables with other exogenous variables. Our main results hold when employing these additional regressions.

10 Because autocorrelation and seasonal autocorrelation are present in the residuals, we introduce an AR structure in the model specification. The structure of residuals is depicted by AR(1) and AR(7) processes. By and large, conclusions are robust to alternative specifications. Visual evaluation of autocorrelation functions (ACF) and partial autocorrelation functions (PACF), analysis of p-values of AR parameters of alternative variants of the baseline specification, and the application of the principle of parsimony to avoid over-fitting led to favouring this specification. Beyond that, there is an expectation of persistence of volatility relative to previous-day events, thereby justifying an AR(1) structure. The AR(7) parameter is explained by the presence of weekday cycles/seasonality.

11 Hereinafter, hatted letters denote point estimates. For instance, ${\hat{\theta }}_2$ is the point estimate of the parameter ${\hat{\theta }}_2$.

12 The EGARCH model is developed in Nelson (1991). Computations are performed in Eviews.

13 Removing seasonal effects directly from the mean equation presents the advantage of reducing measurement-error bias. Indeed, when carrying out a two-step procedure, there is a greater likelihood that measurement error will arise and that some of the effects captured in the second regression will stem from the first stage regression measurement errors.

14 As pointed out by an anonymous referee, the introduction of interactions of D1 with other covariates may raise concerns about multi-collinearity. In non-tabulated analysis, that issue is tackled in the following way. First, we collapse the main sample into two subsets: before and after market coupling entry-into-force. Then, baseline models are re-estimated using the two alternative subsets of data. The results from this supplementary assessment are consistent with the sign and magnitude of estimated parameters found for interactions of $D1$ with other covariates.

15 The average price level over time could turn out to be insufficient to give investors in generation capacity a proper return on their investment.

Additional information

Funding

This work was supported by FEDER/COMPETE [grant numbers POCI-01-0145-FEDER-007659, UID/ECO/04007/2013].