Three-State Sentiment Dynamics
The main objective of this thesis is a detailed modeling of the dynamic components, i.e. transition rates with an embedded 'herding' component, in a stochastic framework. In addition to bold and cautious sentiment neutrality is incorporated into three versions of three-state models of sentiment dynamics. One version is formalized via the transition probability approach while a second version utilizes the discrete choice framework. The third model considers a bold opinion index and a neutrality index. Survey data on American investor sentiment is used to estimate the model parameters via maximum likelihood. The forecasting performance of this model is remarkable in terms of all three sentiments. The models exhibit a realistic description of the underlying group dynamics, including the separate treatment of the neutral agents. The findings provide an indication of the necessary degree of complexity of behavioral models in different scenarios. It can be concluded that the neutrality index adds additional explanatory power.