Finance, Markets and Valuation
Not found
Title: Modelling conditional volatility in the Spanish Ibex-35 stock index using high frequency data. A comparison of the EGARCH model and the backpropagation neural network
Authors: Javier Oliver
DOI: Not found
Abstract:
The analysis of conditional volatility is a necessary step towards the accurate valuation of the risk inherent in financial assets such as stocks, bonds, indices and derivatives among others. A good volatility prediction is required to diversify portfolios, value financial options, calculate VaR, etc. Therefore, it is necessary to generate models that are capable to predict financial assets’ volatility. At present, the most widely employed models are those belonging to the GARCH family. In this paper the conditional volatility of the Spanish IBEX-35 stock index with high frequency data is analyzed by means of an ARMA-EGARCH model, as this model can capture the asymmetries in the index volatility. Next we apply the backpropagation neural network to the same end and compare the results obtained. The comparison is made using the same variables in both models in order to obtain a more balanced and fair comparison. The results show that the neural network is a good alternative to the traditional GARCH family models. In fact, in the analysis, the backpropagation neural network repeatedly beats the ARMA-EGARCH model regardless the frequency of the data and the error measurement type employed.
Keywords: Not found
