Covariance estimation using high-frequency data: Analysis of Nord Pool electricity forward data

Abstract

Volatility and correlation modelling is important in order to calculate hedge ratios, value at risk estimates, CAPM betas, derivate pricing and for risk management in general. Historically, these measures have usually been obtained by analyzing daily data. Recently access to intra-daily high-frequency data for two of the most liquid contracts at the Nord Pool exchange (quarterly and yearly forward contracts), makes it possible to apply new and promising methods for analyzing volatility and correlation. We apply the concept of realized volatility and realized correlation, and as the first study statistically describe the distribution (both distributional properties and temporal dependencies) of electricity forward data from 2005 to 2009. The overall main findings show that the logarithmic realized volatility are approximately normal distributed, while realized correlation seems not. Further, realized volatility and realized correlation has a long memory feature, and there seem to be a high correlation between realized correlation and volatilities. These results are to a large extent consistent with earlier stylized facts studies of other financial and commodity markets.