We use weekly survey data on short-term and medium-term sentiment of German investors to estimate the parameters of a stochastic model of opinion dynamics. The bivariate nature of our data set also allows us to explore the interaction between the two hypothesized opinion formation processes, while consideration of the simultaneous weekly changes of the stock index DAX enables us to study the influence of sentiment on returns within a behavioral model of boundedly rational traders. Technically, we extend the maximum likelihood framework for parameter estimation in agent-based models introduced by Lux (2009a) by generalizing it to bivariate and trivariate settings. As it turns out, short-term sentiment is governed by strong social interaction with abrupt changes of direction while medium-term sentiment is a slowly moving process with more moderate social interaction. The trivariate model can potentially predict stock returns out-of-sample on the base of medium-run sentiment at least if an apparently spurious influence from short-run sentiment is discarded.