Traditional financial theory predicts that comovement in asset returns is due to fundamentals. An alternative view is that of Barberis and Shleifer (2003) and Bar- beris, Shleifer and Wurgler (2005) who propose a sentiment based theory of comovement, delinking it from fundamentals. In their paper they view comovement under the prism of the standard Pearson's correlation measure, implicitly excluding extreme market events, such as the latest financial crisis. Poon, Rockinger and Tawn (2004) have shown that under such events different types of comovement or dependence may co-exist, and make a clear distinction between the four types of dependence: perfect dependent, independent, asymptotically dependent and asymptotically independent. In this paper we extend the sentiment based theory of comovement so as to cover the whole spectrum of dependence, including extreme comovement such as the one that can be observed in financial crises. One of the key contributions of this paper is that it formally proves that assets belonging to the same category comove too much in the tail and reclassifying an asset into a new category raises its tail dependence with that category.
