Gentrification in NYC
Manuel Rueda · December 16, 2015
Manuel Rueda · December 16, 2015
Gentrification is a subject that has received significant media and academic attention in the last few decades, mainly because it’s a process powered by the possibility of mobility which has been limited in earlier years. Given the recent rise of this phenomena, researchers have not yet been able to reach an agreement on what are its causes, its social and economical effects, or even on what constitutes a formal process of gentrification. Regardless of this, the subject is one that continues to gain importance, mainly because it has been accelerating on recent years, and it has significant implications in public policy; for example, in several studies it has been negatively associated with the displacement of population and loss of capital investment.
An interesting study performed on this subject is that of Nathalie Voorhees from Chicago's Center for Neighborhood and Community Improvement. On it, the researcher takes 13 variables that have been associated with gentrification and neighborhood upgrading in the literature and uses them to create a Gentrification Index that evaluates this process through Chicago's different community areas. More details can be found here. We will translate Voorhees' study to the New York Metro area, investigating the relevance of our results under this different social and economic context.
The data utilized on this study is retrieved from the US Census American Community Service as of 2013, and corresponds to the census tract data for the counties of New York (Manhattan & Staten Island), Kings (Brooklyn), Queens and the Bronx. For a couple of tracts where data was not available a spatial interpolation was performed, although the number of affected points is relatively low (~3%). All statistical and spatial anaylsis has been performed using R, GeoDa, and QGIS.
Below we present the variables identified by Voorhees, along with the theoretical relationship they hold with gentrification. By hovering of each of them it is possible to see the abreviation that will be used for each variable throughout the study. The 2 most prominent authors on the field are David Ley and Daniel J. Hammel, with their classical research papers accounting for a large percentage of those available on the subject. A number of other authors have recently contributed to the analysis by proposing a couple of new covariates.
| Ley (1992-99) | Hammel, et al. (1996-2004) | Other autors (2010-2014) |
|---|---|---|
| Median House Value (+) | Pct. Manager Occupations (-) | Pct. Female Led Households (-) |
| Median Family Income (+) | Pct. College Education (-) | Pct. Children (<19 years) (-) |
| Pct. Low Income Families (-) | Pct. Owner Occupied Homes (-) | Pct. Private School Attendance (+) |
| Pct. Eldery (>65 years) (-) | Pct. White (-) | |
| Pct. Black (-) | ||
| Pct. Latino (-) |
It would be interesting to know how each of the variables interact with each other. If the theory is correct, we would expect those positively associated with gentrification to be positively correlated among each other. We would also expect a negative correlation with those variables negatively associated to gentrification. The chart below help us test these assumptions by showing a correlation matrix among all variables.
We see that in general the expected relationship holds, although reading the graph becomes dificult due to the large amount of data in display. Now we will take another approach: given that gentrification is directly and indisputably associated with increases in Home Values, and considering this is probably the most tangible effect in terms of public policy (i.e. the main driver of displacement), we will use HOUSEVAL13 as our independent variable and investigate the correlation of all other variables against this one. To make the analysis more granular, we will also split the data per county. Below we can find the results.
Some interesting results arise: although the direction of association is in line with our expectations (or practically null), the magnitudes differ significantly accros counties. We see the largest interactions occuring on New York and Kings, while they are severely reduced on the Bronx. This suggests that the gentrification phenomena is more prevalent on these first two mentioned regions.
In order to build the gentrification index, we will use the comparative approach followed by Voorhees. For it, we will look at the level of each variable on each census tracts, and compare it to the mean of the population (in this case the mean of the county, to account for possible structural differences among them). If the value is above-average, and the theoretical association with gentrification is possitive, we assign the tract a score of +1. If the proposed association is negative, then the tract receives a score of -1. Following the same logic we also assign scores to the below-average tracts. At the end the scores for all variables are added up, and those tracts with positive values are identified as "getrifying".
The first map below presents the distribution of 3 variables making up the index: Pct. College Education, Pct. Black and Median House Value, along with the actual Gentrification Index. Other variables were excluded due to limitations with CartoDB Academic Licence.
At first glance we can see a clustering of gentrification-contributing variables on Mid and Lower Manhattan, south-west Brooklyn, east of Queens, and north of the Bronx, although these last two appear less prevalent. To continue with the analysis, now we present Statistically Significant Clusters (95%), which are derived using the Moran's I indicator. This should give us more confidence on our findings. To switch across variables use the same menu from the map above.
VARIABLE SELECTOR
The clustering analysis confirms our findings, revealing that these are statistically relevant. We also see negative clusters occuring on the upper-east Brooklyn and the South of the Bronx, areas that are know to have adverse economic conditions.