Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Forecasting Swedish Stock Market Volatility and Value-at-Risk: A Comparison of EWMA and GARCH Models

Nilsson, Carl LU (2017) NEKP01 20171
Department of Economics
Abstract (Swedish)
In this study we compare different volatility models on their ability to forecast one day ahead volatility and value-at-risk (VaR). We compare five different GARCH specifications: GARCH, IGARCH, GJR-GARCH, EGARCH and APARCH, as well as EWMA, each paired with six different conditional distributions. These models are used to forecast volatility and VaR one day ahead using daily return data from the Swedish stock market index OMXS30. The forecasts are then compared using the model confidence set procedure of Peter Reinhard Hansen, Asger Lunde, and James M Nason (2011). “The model confidence set.” In: Econometrica 79.2, pp. 453–497.

We find the APARCH models best for forecasting volatility, while for forecasting VaR the best models are... (More)
In this study we compare different volatility models on their ability to forecast one day ahead volatility and value-at-risk (VaR). We compare five different GARCH specifications: GARCH, IGARCH, GJR-GARCH, EGARCH and APARCH, as well as EWMA, each paired with six different conditional distributions. These models are used to forecast volatility and VaR one day ahead using daily return data from the Swedish stock market index OMXS30. The forecasts are then compared using the model confidence set procedure of Peter Reinhard Hansen, Asger Lunde, and James M Nason (2011). “The model confidence set.” In: Econometrica 79.2, pp. 453–497.

We find the APARCH models best for forecasting volatility, while for forecasting VaR the best models are either APARCH, GJR-GARCH or EGARCH—depending on which level of VaR we use—paired with conditional distributions that take skewness and excess kurtosis into account. EWMA, GARCH and IGARCH specifications cannot be recommended either for forecasting volatility or for forecasting VaR. (Less)
Please use this url to cite or link to this publication:
author
Nilsson, Carl LU
supervisor
organization
course
NEKP01 20171
year
type
H2 - Master's Degree (Two Years)
subject
keywords
volatility forecasting, VaR, GARCH, model confidence set
language
English
id
8923808
date added to LUP
2017-09-12 11:52:15
date last changed
2018-07-02 13:01:33
@misc{8923808,
  abstract     = {{In this study we compare different volatility models on their ability to forecast one day ahead volatility and value-at-risk (VaR). We compare five different GARCH specifications: GARCH, IGARCH, GJR-GARCH, EGARCH and APARCH, as well as EWMA, each paired with six different conditional distributions. These models are used to forecast volatility and VaR one day ahead using daily return data from the Swedish stock market index OMXS30. The forecasts are then compared using the model confidence set procedure of Peter Reinhard Hansen, Asger Lunde, and James M Nason (2011). “The model confidence set.” In: Econometrica 79.2, pp. 453–497.

We find the APARCH models best for forecasting volatility, while for forecasting VaR the best models are either APARCH, GJR-GARCH or EGARCH—depending on which level of VaR we use—paired with conditional distributions that take skewness and excess kurtosis into account. EWMA, GARCH and IGARCH specifications cannot be recommended either for forecasting volatility or for forecasting VaR.}},
  author       = {{Nilsson, Carl}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Forecasting Swedish Stock Market Volatility and Value-at-Risk: A Comparison of EWMA and GARCH Models}},
  year         = {{2017}},
}