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), then shows how this auxiliary data is economically meaningful.

This paper provides empirical evidence on how expectation errors of forecasted and realized credit spreads can predict future macroeconomic outcomes. It first predicts a series of past credit spread expectations by applying textual analysis through natural language processing and then uses the textual factors in predictive regressions spanning 1948 to 2022.

I study the impact of various measures of monetary policy announcements on household expectations between 2013 and 2021. I approach this through an event study and a local projections methodology that exploits the microdata found in the Survey of Consumer Expectations as well as the timing of the FOMC meeting announcements in this time period.