Abstract
This paper employs Latent Dirichlet Analysis for Survey Data (LDA-S) to identify and classify households into distinct belief types based on their responses in the Survey of Consumer Expectations (SCE). I uncover three belief types – inconsistent/uncertain, pessimistic, and optimistic – characterized by unique patterns of expectations about macroeconomic and personal financial conditions. I demonstrate that incorporating these belief types into a simple static model significantly improves the explanatory power for predicting households' 12-month inflation expectations. My findings indicate that households with pessimistic beliefs tend to have higher inflation expectations, while those with optimistic beliefs have lower expectations. These results emphasize the importance of central banks considering the type of information households consume and tailoring their communication strategies accordingly. Understanding the nature of these belief types and their impact on economic behavior can lead to more effective policy measures and better-targeted communication strategies. Moreover, this research highlights the potential of using latent class analysis techniques to extract valuable information from survey data, which can be applied in various economic contexts.
