Estimation and Analysis of VaR on forwards' data in Nordic Electricity Market
(2012) NEKN02 20121Department of Economics
- Abstract
- It was intended with this research to estimate in-sample and out-of-sample Value-at-Risk(VaR) on forwards’ data in Nordic electricity market by using different parametric and nonparametric approaches and investigate which of them give the most accurate results after implementation of back-testing. There were two confidence levels of 95% and 99% used in this research.
After implementation of data analysis; the following approaches were chosen for calculation of VaR: basic HS, age-weighted HS, volatility-weighted HS, normal and t-distribution approaches. GARCH and EGARCH models were run with both normally and t-distributed innovations in order to account for volatility clustering.
The results obtained indicated that basic HS approach... (More) - It was intended with this research to estimate in-sample and out-of-sample Value-at-Risk(VaR) on forwards’ data in Nordic electricity market by using different parametric and nonparametric approaches and investigate which of them give the most accurate results after implementation of back-testing. There were two confidence levels of 95% and 99% used in this research.
After implementation of data analysis; the following approaches were chosen for calculation of VaR: basic HS, age-weighted HS, volatility-weighted HS, normal and t-distribution approaches. GARCH and EGARCH models were run with both normally and t-distributed innovations in order to account for volatility clustering.
The results obtained indicated that basic HS approach performed very well for all forwards with different delivery periods for estimation of both in-sample and out-of-sample VaR. However, this result should be considered with caution; as there is a threat of overestimation of risk noticed. Age-weighted HS approach applied for in-sample VaR underestimated the risk for quarterly and yearly forwards. Normal distribution approach accounting for volatility clustering also gave accurate results in every case for both in-sample and out-of-sample VaR. The relevance of taking different confidence levels was proven by the fact that t-distribution approach worked mostly when using 99%. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/2701867
- author
- Streimikyte, Egle LU and Grainyte, Indre LU
- supervisor
-
- Rikard Green LU
- Karl Larsson LU
- organization
- course
- NEKN02 20121
- year
- 2012
- type
- H1 - Master's Degree (One Year)
- subject
- keywords
- Value-at-Risk, parametric and non-parametric approaches, back-testing, Christoffersen test.
- language
- English
- id
- 2701867
- date added to LUP
- 2012-06-08 14:34:18
- date last changed
- 2012-06-08 14:34:18
@misc{2701867,
abstract = {{It was intended with this research to estimate in-sample and out-of-sample Value-at-Risk(VaR) on forwards’ data in Nordic electricity market by using different parametric and nonparametric approaches and investigate which of them give the most accurate results after implementation of back-testing. There were two confidence levels of 95% and 99% used in this research.
After implementation of data analysis; the following approaches were chosen for calculation of VaR: basic HS, age-weighted HS, volatility-weighted HS, normal and t-distribution approaches. GARCH and EGARCH models were run with both normally and t-distributed innovations in order to account for volatility clustering.
The results obtained indicated that basic HS approach performed very well for all forwards with different delivery periods for estimation of both in-sample and out-of-sample VaR. However, this result should be considered with caution; as there is a threat of overestimation of risk noticed. Age-weighted HS approach applied for in-sample VaR underestimated the risk for quarterly and yearly forwards. Normal distribution approach accounting for volatility clustering also gave accurate results in every case for both in-sample and out-of-sample VaR. The relevance of taking different confidence levels was proven by the fact that t-distribution approach worked mostly when using 99%.}},
author = {{Streimikyte, Egle and Grainyte, Indre}},
language = {{eng}},
note = {{Student Paper}},
title = {{Estimation and Analysis of VaR on forwards' data in Nordic Electricity Market}},
year = {{2012}},
}