Investor Sentiment Connectedness: Evidence from Linear and Nonlinear Causality Approaches

This paper provides a novel perspective in determining the causality of sentiment across US, Latin America, Eurozone, Japan and Asia (excluding Japan), based on monthly data covering the period of January 2003 to November 2017. Using a survey-based sentiment index of ‘sentix’, our results tend to suggest strong evidence of nonlinearity and structural breaks making the results from linear causality models unreliable. Using a kernel-based multivariate nonlinear causality test, we find that causality runs from Eurozone to US, Asia, and Japan, with Japan also causing the Eurozone sentiment, and Latin-America causing Japanese sentiment. Interestingly, when we applied rolling estimations to detect time-varying causality for the cases of Eurozone and US, Eurozone and Asia, Eurozone and Japan, and Latin-America and Japan, we found evidence of bi-directional spillovers during certain months of the recent global financial crisis, and thereafter. Overall, our findings indicate that the sentiments of Japan, Asia, and US are related quite strongly with that of the Eurozone, as is Japan and Latin America.