Modelling interdependence in a pair of heating oil and natural gas futures curves

Abstract

This Master thesis further develops the framework for the joint modelling of several futures curves proposed by Ohana (2010). The key innovation is the incorporation of dynamic conditional correlation models based on hierarchical Archimedian copula (HAC-DCC). The conducted analysis allowed to forecast the distribution of the returns of any portfolios composed of the available futures contracts for short time periods. As shown in the study, value-at-risk estimates derived from the forecasts produced by HAC-DCC models are accurate, and these models outperform other benchmark models on a consistent basis as shown by the value-at-risk backtesting procedure carried out on a set of 1000 simulated futures portfolios.