Sentiment proxies: Computing market volatility

Abstract

Macroeconomic announcements can have an influential effect on the price, and related volatility, of an object traded in financial markets. Modeling the impact of a relevant announcement on a specific commodity is of interest in building financial models of such objects. The announcements may generate false hopes or correctly indicate the ways in which the prices of objects may change. We describe a bootstrap model which is an attempt to analyse the impact of a publicly released oil inventory announcement on the price of oil futures contracts. A comparison with traditional econometric regression model is presented and we perturb the traditional models with: (a) a dummy time series that contains the dates on which the announcement is made, and (b) a sentiment time series that reflects the sentiment of the market. The sentiment time series is generated using natural language processing techniques. © 2012 Springer-Verlag.