How have varying demographic and housing characteristics of neighborhoods changed relative to up-zoned areas in Brooklyn, New York from 2013-2019?
Background
With the housing crisis New York City presently faces, policymakers have been exploring proactive approaches in using rezoning as a measure to encourage additional development and subsequently increase housing supply within the City. Up-zonings in particular are critical to this process as they increase the overall allowable bulk, density, and size of developments in certain areas beyond what the existing zoning designations and regulations permit. In 2007, the Bloomberg administration introduced Pia NYC 2030, a long-term comprehensive plan for the City, which notably included the goal of creating housing for more than one million New Yorkers. The plan specifically looked to rezoning as a tool to increase the housing stock and affordability in the City. This legacy of up-zoning continued well into De Blasi o’s administration with the East New York Neighborhood Plan, the first major rezoning during that era. However, up-zonings have not always been impactful nor equitable as this process has been throught to show implications across various aspects including but not limited to changes in terms of demographic compositions of neighborhoods, gentrification, and displacement. As such, potential trends like these present a unique opportunity to study up-zoning through a deeper spatial lens by investigating this phenomenon across varying demographic and built environment characteristics to determine which of these display stronger correlations to the changes brought about by up-zoning. Additionally, it allows the very questioning whether or not up-zoning can actually be attributed for its associated benefits along with varying indicators of neighborhood change.

Variables for Analysis
Through an iterative process, this research sought to understand the correlation of chosen demographic and built environment variables to the process of “up-zoning” from 2013-2019. Up-zoning, as a clear and defined variable, requires qualitative review to understand which indicators might produce a model of best fit. Our initial Ordinary Least Squares Regression found that of the 6 variables initially chosen to indicate an upzoning- Changes in Residential FAR, Built FAR, New Residential Units, Built Floor Area, Assessed Land Value, Building Height, Distance from the Nearest Upzoning- Assessed Land Value and Distance from the Nearest Up-zoned Areas, failed as a result of multicollinearity when tested in the Geographically Weighted Regression model. Additionally, these up-zoning variables would be studied against a plethora of dependent variables classified between a total of 52 demographic and housing characteristic changes based on relevant literature during the study period.

Methodology
The methodological approach to investigate the correlations between associated up-zoning variables and the corresponding demographic and housing characteristics was to first prepare the data of each variable by calculating their percent changes during the study period. Next, multiple iterations and combinations of identified independent variables relevant to the indicators of up-zoning were tested using Ordinary Least Squares Regression Modeling to determing the best-fit model to describe up-zoning as a phenomenon. In this phase, variables were tested narrowed down based on collinearity as well. Then, individual Geographically Weighted Regressions were conducted for all 52 dependent variables across all of Brooklyn using the best-fit model. A distance band of 2 miles was used to be able to capture and study the relationship of up-zonings and the dependent variables on a neighborhood scale. Finally, to contextualize the findings, the results were analyzed against the historical boundaries and surrounding half-mile and 1-mile buffers around up-zoned areas during the study period to determined whether of not stronger relationships are located within or in surrounding areas of these boundaries. The resulting GWRs with the top 3 highest R’ values for demographic and housing characteristic changes were then studied further. The following diagram outlines the general steps taken along with the main purpose of each section: