TY - THES ID - 146381766 TI - Forecasting S&P500 volatility by characterizing shocks using Latent Semantic Analysis on new articles AU - Moreno Miranda, Nicolas AU - Lambert, Marie AU - Ittoo, Ashwin AU - Platania, Federico PY - 2016 PB - Liège Université de Liège (ULiège) DB - UniCat KW - LSA KW - GARCH KW - EGARCH KW - GARCH-X KW - Latent Semantic Analysis KW - Volatility Forecasting KW - S&P500 KW - Lagged corredlations KW - Reuters KW - News KW - News Articles KW - Conditional Volatility KW - Sciences économiques & de gestion > Finance UR - https://www.unicat.be/uniCat?func=search&query=sysid:146381766 AB - The information contained in news articles plays a key role on
financial markets. It may describe changes in the fundamentals of a company
or influence the way investors perceive the risk associated with it. This paper
aims at measuring with mathematical means the main underlying semantic
content of news articles, such that it captures information useful to forecast
volatility.

A modified EGARCH model with external factors, obtained from a latent semantic alaysis on news articles, is proposed to measure the impact on volatility induced by the latent semantic content of the textual news data. I find that several semantic dimensions play an important
role in explaining observed volatility, while others are useful to forecast it. It
is likely, that with further research, a model based on semantic content could
greatly improve our understanding of the market’s response to news releases. ER -