Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Bivariate VaR Estimation in energy forwards – An Extreme Value Approach with Copulas and GARCH Volatility

Thorselius, Elsa LU (2013) EXTM10 20131
Department of Economics
Abstract
The overall goal of this thesis is to examine the effect specific model components, GARCH, EVT and copula, have when examining the possible losses that can occur when holding a portfolio of financial energy contracts within different markets. Daily closing prices of financial monthly forward contracts from 2008-2012 are used in this study. The contracts are in the Nordic and German power markets and seven different combinations of contracts were used for examining the possible losses. Using the actual return series, Monte Carlo simulations in a model that consists of GARCH estimation for modelling the non-constant volatility of the return series, Extreme Value Theory using the Generalized Pareto Distribution for examining the tail... (More)
The overall goal of this thesis is to examine the effect specific model components, GARCH, EVT and copula, have when examining the possible losses that can occur when holding a portfolio of financial energy contracts within different markets. Daily closing prices of financial monthly forward contracts from 2008-2012 are used in this study. The contracts are in the Nordic and German power markets and seven different combinations of contracts were used for examining the possible losses. Using the actual return series, Monte Carlo simulations in a model that consists of GARCH estimation for modelling the non-constant volatility of the return series, Extreme Value Theory using the Generalized Pareto Distribution for examining the tail behaviour of the distribution and a copula function in order to capture the dependence between the two return series was performed. In order to obtain the potential losses Value at Risk (VaR) was used on the final simulated return series. This thesis finds the impact of the different components and especially the significance of including the dependence with a copula function to vary with the chosen significance level. For a low significance level the difference when excluding the dependence structure is more significant. The effect also varied with the correlation between the two contracts used in a portfolio. When two highly correlated contracts were used the difference was larger when excluding the dependence structure than for less correlated combinations. (Less)
Please use this url to cite or link to this publication:
author
Thorselius, Elsa LU
supervisor
organization
course
EXTM10 20131
year
type
H2 - Master's Degree (Two Years)
subject
keywords
GARCH, Extreme Value Theory, Copula, Energy Forwards, Nordic Power, German Power, Finance, t-distribution, t-copula, Value at Risk, Monte Carlo, Generalized Pareto Distribution
language
English
id
4193981
date added to LUP
2013-12-10 10:25:00
date last changed
2013-12-10 10:25:00
@misc{4193981,
  abstract     = {{The overall goal of this thesis is to examine the effect specific model components, GARCH, EVT and copula, have when examining the possible losses that can occur when holding a portfolio of financial energy contracts within different markets. Daily closing prices of financial monthly forward contracts from 2008-2012 are used in this study. The contracts are in the Nordic and German power markets and seven different combinations of contracts were used for examining the possible losses. Using the actual return series, Monte Carlo simulations in a model that consists of GARCH estimation for modelling the non-constant volatility of the return series, Extreme Value Theory using the Generalized Pareto Distribution for examining the tail behaviour of the distribution and a copula function in order to capture the dependence between the two return series was performed. In order to obtain the potential losses Value at Risk (VaR) was used on the final simulated return series. This thesis finds the impact of the different components and especially the significance of including the dependence with a copula function to vary with the chosen significance level. For a low significance level the difference when excluding the dependence structure is more significant. The effect also varied with the correlation between the two contracts used in a portfolio. When two highly correlated contracts were used the difference was larger when excluding the dependence structure than for less correlated combinations.}},
  author       = {{Thorselius, Elsa}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Bivariate VaR Estimation in energy forwards – An Extreme Value Approach with Copulas and GARCH Volatility}},
  year         = {{2013}},
}