A small area index of gentrification, with application to New York City

Keywords: Gentrification, Gentrification Index, Principal Components Analysis, Bayesian Conditional Autoregressive Model, New York City, Home value, Housing cost
Session Type: Virtual Paper Abstract
Day: Monday
Session Start / End Time: 2/28/2022 05:20 PM (Eastern Time (US & Canada)) - 2/28/2022 06:40 PM (Eastern Time (US & Canada))
Room: Virtual 72

Abstract

A small-area index of the multi-factorial phenomenon loosely referred to as gentrification is presented, with application to New York City (NYC).

We first identified five separate variables that represent the nuanced and processual nature of the socioeconomic and demographic shifts associated with gentrification (median family income, median rent and proportions of non-Hispanic white, 20-34-year-olds and adults with a 4-year college degree). The relative change of these variables was computed from the years 2000 to 2016 for NYC census tracts that are spatially normalized to the year 2010. Raw scores derived from principal components analysis of suitably transformed input variables were then spatially smoothed through a fully Bayesian conditional autoregressive model, presenting an innovation over previous methods since it essentially blurs the otherwise artificial tract boundaries and results in a simulated posterior distribution of scores for each tract. The median and upper/lower percentiles provide a point estimate and assessment of uncertainty, respectively, of gentrification for each tract.

The composite index is positively associated with census tract-level change in home values, as measured separately from a real property database. Furthermore, the index is strongly spatially associated with brownfield projects and “upzonings” during the same period that the index applies to. While our motivation is understanding the forces and dynamics of gentrification in New York City (NYC), the methodological approach is expected to be replicable in other urban areas using readily-available public domain data similar to the US census and American Community Survey.