Spatial Analysis

Exploring the Relationship Between Neighborhoods and Concentrated Disadvantage in Baltimore City, Maryland

This project explores the distribution and potential factors affecting racial residential segregation in Baltimore. Through the use of spatial visualization, spatial lag models to assess spatial autocorrelation, historical redlining data to predict settlement patterns, and identification of gentrified areas, a comprehensive look at the ethno-racial composition of this city becomes possible.

Distribution of White Residents

This visualization shows the distribution of residents who identify as white throughout the city. This exploration began by first assessing if there seems to be spatial autocorrelation within the settlement patterns of residents of different racial groups. Plotting percent white serves as a proxy for capturing diversity such that majority white areas are less diverse compared to areas that are predominantly non-white. In this map, it is clear that majority white spaces (the darkest shaded regions) do appear to be clustered near each other, specifically in the northwest of the city and in the south east near the harbor.

Distribution of Poverty

This visualization represents the spatial distribution of residents living in poverty across the city. Plotting poverty allows us to see which areas are the most impoverished and likely experiencing concentrated disadvantage. It seems that the poorest parts of the city are in the middle and span outwards in most directions from there. What is most interesting is comparing the poorest areas to the map of percent white residents to the left. It looks like the predominantly white areas have the lowest rates of poverty, while the areas that are the least white look to have the highest rates of poverty. Between these two plots, it does seem that there is preliminary evidence of racial residential segregation and primarily non-white areas are the most impoverished.

Redlining Data and Distribution of White Residents

Above is a density plot of the distribution of white residents across the city in 2010 disaggregated by which letter was assigned to respective census tracts when redlining was actively occurring. Typically, when looking at how redining has affected racial composition of neighborhoods over time, we would expect to see that tracts graded as a C or D would be predominantly non-white areas because legacies of historic racial discrimination and segregation often have ling-lasting effects. In this type of plot specficially, we would expect to see much less overlap of the different grades and the green and purple distributions would tend to have the lowest percentage of white residents. However, this plot presents us with quite striking data. Areas that were graded as C have the highest percentage of white residents. Further, there is quite a bit of overlap between the different distributions seeming to indicate that historic redlining categories are not very predictive of racial composition of neighborhoods in 2010. As this is a surprising result, some other social process must be at work that is altering where we would expect to see higher versus lower percentages of white residents across the city.

Past and Present Poverty and Racial Composition

This scatterplot shows how census tracts have changed over the course of 9 years with respect to poverty rates. Also, the points are colored by how each tract has changed regarding racial composition of precentage of white residents in the tract. This plots allows us to see if there are some tracts are becoming more or less impoverished and also to what extent they are changing in racial composition. After considering what the redlining data predicted about racial composition, I theorized that gentrification could be at work in Baltimore, with usually more white residents moving into majority non-white and poorer areas as they are reformed and bring in new businesses. Thus, gentrification could explain the pattern observed in the plot to the left. In this plot, the points that are of most interest are those that have lower poverty rates since 2010 and higher percentages of white residents, as this would indicate the typical result of gentirifcation. These points are located in the lower left quadrant are colored purple. Also of interest from a policy perspective are the points that have lowered poverty since 2010 but have maintained their racial diversity; these would also be in the lower left quadrant colored in shades of orange/red.


This course and this project introduced me to a large amount of new material about how social factors and population characteristics are distributed across space. Most important though, was learning how neighborhoods affect characteristics of other nearby areas and how neighborhoods are extremely influential in an individual's life chance. Using these various tools to perform spatial analysis truly allows data scientists and sociologists to examine how things like ethno-racial composition and poverty are distributed spatially and if they appear to influence the racial composition or poverty in neighboring tracts.

Using tools like chloropleth maps, spatial lag models, spatial Durbin models, and OLS regressions to assess how characteristics have changed over time, a story emerges in the city of Baltimore about the degree to which it is racially residentially segregated. These tools help to unpack this initial observation by seeing if historic redlining practices are predictive of the residential segregation that is present. And when this was not the case, regressions using different points in time and residual change allowed for evaluation of potential gentrification occurring in the city that could provide insight on some of the observed settlement patterns. Overall, these tools are exceptionally useful for city and local governments who hope to identify areas of their cities that would benefit most from policy interventions, as well as trying to see what types of social programming would be most useful across their jurisdictions.