An empirical study of the Value-at-Risk of the renewable energy market and the impact of the oil price

Renewable energy is gaining increasing importance in the generation of power due to the finite existence of fossil fuels and concerns about climate change. As its demand grows financial interest from investors’ increases, thus it is important to find the most effective way of quantifying the risk of the renewable energy market. Furthermore as renewable energy can be viewed as an economic substitute for other energy sources such as crude oil - a commodity that has been known to have a significant impact on financial markets - an empirical relationship is likely to exist between the two resources. This paper will assess the best way of measuring the risk of the renewable energy market by using one of the most common risk measurement tools Value-at-Risk. Using daily data of the return observations of five renewable energy indices between the 1st of January 2004 and the 12th of June 2015 a total of 2987 observations, the VaR will be estimated for each of these indices. This is achieved using both parametric and non-parametric methods, and then backtesting these using the two-sided Kupiec test to determine which method provides the best estimate of VaR. The non-parametric methods employed in this paper are the Basic Historical Simulation (BHS) and the Exponentially Weighted Moving Average (EWMA) model. The parametric methods applied are the Generalized Autoregressive Conditional Heteroskedasticity, or GARCH (1, 1) model and the Threshold-GARCH, or TGARCH, using both the normal and Student-t distribution. The sample period is split into an in-sample period of 522 days and an out-of-period of 2465 days, where the 522 days will be used as the size of the “rolling-window” which is used to calculate the VaR throughout this paper. After determining which model provides the best estimate of VaR a regression will be run using this VaR estimate as the dependent variable, and the oil price and the three-month rate of a US Treasury bill - taken as the interest rate – as the explanatory variables. The results show that the parametric methods outperform the non-parametric methods with the GARCH (1, 1) model under the Student-t distribution in particular providing the best estimate of VaR. In general they show that the models which can account for heavy-tailed distributions perform better, with all models using the Student-t distribution giving better estimates than the normal distribution. Furthermore a statistically significant relationship between the VaR estimate of any given renewable energy index and the oil price was identified, with a rise in the price of oil causing a decrease in the VaR estimate of the given renewable energy index