Ambiguous Text

Text is inherently ambiguous. I use textual analysis to extract beliefs from financial news and social media sources to construct a Big Data ambiguity index. The index uses mixture of distributions that regards disagreement between sources and their individual uncertainties. Empirical results show the ambiguity index as an omitted variable bias in the fundamental relation between risks and returns. An ambiguity averse agent is shown to love ambiguity in unfavorable markets and dislike ambiguity in favorable markets. The paper finds that news media have a bigger influence on asset prices than social media except during the recession from Jun 2009 to Nov 2016