CASA Seminar Series: Elizabeth Delmelle, UNC Charlotte

Elizabeth Delmelle, UNC Charlotte: Spatial Variations in Property Advertisement Text

About this event

Link

In this presentation, I present two case studies that use text analysis and natural language processing techniques to understand the real estate and rental landscape according to how properties are advertised. In the first, we explore how the language used to market single family owner-occupied dwellings varies by the racial and income makeup of neighborhoods. We begin by classifying neighborhoods in Charlotte, North Carolina (USA) according to trends in the race and income of mortgage home applicants over time using a latent growth curve model. Then, using a sample of property listing data, we compare words used to advertise properties for these different types of neighborhoods. Our results show that neighborhoods with largely minority homebuyers are less likely to mention schools and the specific name of the neighborhood. They were also less likely to mention terms associated with capital investment (build, builders, teardown) as compared to Whiter and wealthier neighborhoods. Neighborhoods undergoing processes of gentrification were also unlikely to mention schools, but instead referred to proximity to light rail and the urban core.