In order to inform model building, we first created choropleths of our four predictors (percentage of ESL speakers, percentage of non-White residents, rent burden, and population density) with eviction rates overlaid in Brooklyn by census tract in 2010.
Census tracts with a higher percentage of ESL speakers appear to have lower eviction rates. English language nativity appears to be geographically clustered.
Census tracts with a higher percentage of non-White residents appear to have higher eviction rates. Racial composition of census tracts appears to be geographically clustered.
The relationship between eviction rates and rent burden were not immediately obvious. However, census tracts with high rent burdens appeared to be geographically clustered.
A weak relationship appears to exist between population density and eviction rates, with eviction rates lower in more densely populated census tracts. Census tracts with high rent burdens appeared to be weakly geographically clustered.
Based on our exploratory data visualizations, we identifed significant geographical clustering of our outcome of interest, eviction rates, as well as our predictors of interest (percentage of ESL speakers, percentage of non-White residents by census tract, rent burden, and population density).
With regards to the relationship between our hypothesized predictors and eviction rates, we identified a clear positive association between the percentage of non-White residents and evictions rates in a given census tract. The percentage of ESL speakers in a given census tract appeared to be inversely related to eviction rates; we decided to further examine whether the percentage of non-White residents confounded the relationship between percentage of ESL speakers and eviction rates. Relationships between rent burden, population density, and eviction rates were less obvious; we conducted further exploratory analyses during model building.
Gloria Hu, Naama Kipperman, Will Simmons, Frances Williams
Visualizations and analyses performed using R (v3.6.1) and RStudio (v1.2.1335).
Additional interactivity added using plotly (v4.9.0) and Shiny (v1.3.2).
Click here to see details on all programs used.
2019