News‐based measures of real estate market uncertainty

Using semantic analysis of newspaper content, we introduce a novel index that reflects the public perception of real estate uncertainty ( NewsREU ). The NewsREU is model‐free and offers straightforward interpretability. We benchmark the NewsREU against a model‐based real estate uncertainty ( REU ) measure and show that the NewsREU contains incremental information about the real estate sector beyond what the REU captures. Moreover, we utilize an advanced topic model, leveraging neural inference network and contextual embeddings, to effectively unveil the underlying themes in narratives surrounding real estate uncertainty. This approach allows for decomposition of the NewsREU into distinct Topic NewsREU components that may reflect different facets of uncertainty within the realm of real estate. We find that different types of real estate uncertainty may either positively or negatively forecast future housing outcomes. These results underscore the distinct contribution of our real estate uncertainty indices in advancing our understanding of the role of real estate uncertainty.