In their own words: A mixed-methods exploration of public perceptions of affordable housing and their connections to support

Abstract

One of the main barriers faced by proposed affordable housing developments is local public opposition. Consequently, this exploratory study seeks: 1) to determine the U.S. public's current top-of-mind perceptions of affordable housing, and 2) to understand which of these perceptions are related to self-reported support for affordable housing in one's own neighborhood. We employ a mixed-methods approach, administering a nation-wide online survey (N = 534) of close- and open-ended questions. While no majority understood definition emerged, through our discursive analysis and natural language process (NLP) topic modeling we uncover common and persisting perceptions of current affordable housing buildings as federally supported apartments that are run-down and in undesirable neighborhoods. Narratives about residents are limited, though when prompted participants have a clear perceived racial profile of residents. Using a conditional inference regression tree (CI-Tree) and NLP language feature associations, we also find that perceptions focused on government involvement, subsidies, and unsafe neighborhoods are significant predictors of lower support for proposed developments. Alternatively, mentioning the financial aspects of affordable housing is significantly associated with higher support. These findings demonstrate the potential of machine learning methods in uncovering pathways to support affordable housing and can be leveraged for effective framing of proposed developments.