Modelling day-ahead Nord Pool forward-price volatility: realized rolatility versus GARCH models

Abstract

The argument that better volatility estimates can be obtained by using standard time-series techniques on non-parametric volatility measures constructed from high- frequency intradaily returns has been prevalent over the past decade. This study uses high- frequency data and the concept of realized volatility to make one-day-ahead predictions of Nord Pool forward-price volatility. We compare the predictions obtained from realized volatility using standard time-series techniques with the more traditional GARCH framework. Additionally, we examine whether different approaches of decomposing the total variation, and whether inclusion of exogenous effects, improves the accuracy or not. The main findings suggest that significant improvements in the one-day-ahead Nord Pool forward-price volatility predictions can be obtained by applying high-frequency data and the concept of realized volatility.