CFE 2015: Start Registration
View Submission - CFE
A0686
Title: Theory and methods in using text data for economic prediction: The UCL relative sentiment shift approach Authors:  David Tuckett - UCL (United Kingdom) [presenting]
Abstract: Directed Algorithmic Text Analysis is introduced. Its usefulness is illustrated for nowcasting and forecasting the economy and for assessing financial fragility. The method relies on Conviction Narrative Theory, a social-psychological approach to decision-making under uncertainty which focuses the analysis of text documents on specific emotions to detect relative sentiment shifts - RSS. Using Reuters News articles for 1996-2015 RSS out-performs current forecasting methods for the US economy.