Geographic Information Systems Methods

How did we determine populations within Special Emphasis Neighborhoods and the buffer around Unity Park?

Special Emphasis Neighborhoods (SENs) in Greenville don’t neatly align with standard census boundaries. For example, in the map below, the shaded polygons show the SENs for Southernside, West End, and West Greenville.

As the map shows, census tract 7 contains the southern portion of Southernside, all of West End, and the eastern portion of West Greenville. When the Census Bureau reports data for that tract, information from each neighborhood, or portions thereof, are included. Therefore, when the Census Bureau reports median household income for tract 7 ($37,604 for 2019), the number is derived from West End, in its entirety, and portions of Southernside and West Greenville.

In order to fit the census data to local boundaries, like individual SENs, more sophisticated tools are required. These tools fall under the umbrella of Geographic Information Systems (GIS). To ensure accuracy, we utilized information available at the smallest geography possible, census blocks. Census block groups are larger and comprised of these blocks. Census tracts are comprised of block groups. In urban areas, census blocks roughly correspond to a city block. In more rural areas, blocks typically cover more geography due to lower population densities. In these areas, blocks are often bounded by roads or natural features like lakes and streams. Across Greenville County, the median total population for a census block in 2020 was 27 residents, so blocks are a fairly granular measure of population. For population data, blocks are more precise.

Block--level data is reported for the decennial census, which is a full count, rather than a sample measure. To protect privacy, only certain measures are reported by block. These measures include age, race, population (including population in group quarters), and housing units (total, occupied, and vacant). Income is not reported at the block-level.

For this study, population and race were determined during decennial census years for each SEN using the centroid containment method ( see footnote 1 ). Census blocks were selected as being part of an SEN if the true centroid, or geometrical center, was contained by the SEN boundary. Population and race data were then tabulated for all of the contained blocks.

In the example above, the SEN of Southernside is shown in purple. The census blocks that meet the criteria above are shown in light blue. This is a much better fit to local geography than tracts. Census blocks that do not meet the centroid selection criteria are indicated in light brown and were not included in the Southernside population and race tabulation. This methodology was followed for each of the thirteen SENs, as well as calculations for the 1-mile Unity Park buffer.

Population estimates using the true centroid containment method were consistent with other aggregation methods, including areal interpolation ( see footnote 1 ). The latter method weights population based on the proportion of the tract contained within the custom boundary—in this case the SEN or the 1-mile buffer. True centroid containment measures differed from areal interpolation measures by a very small amount (.05%) for total population, .7% for black population, and 1.0% for white population for 1-mile buffer calculations.

One particular SEN, Brutontown, did show differences based on how the centroid was defined. Population and race values for Brutontown should be interpreted with these differences in mind. The example below is a special case in which these factors significantly affect total values. The other 12 SENs were not significantly impacted by inclusion criteria or block boundary changes over time.

Census block 124 (1990) is included in the Brutontown tabulation when the true centroid method is applied. The centroid falls within the Brutontown boundary, and therefore, block 124 populations are included. The block population of this particular block is relatively high (412 residents vs. a mean value of 27). The relatively high number of residents is likely due high population density in the northeast portion of the block. When applying different criteria, which requires that the centroid be placed within this uniquely shaped polygon, the centroid lies just outside of the Brutontown boundary and would not be included. This is a unique case in our analysis where slightly different centroid definitions, the proximity of the centroid to the SEN boundary, and a block with a large population impacts population numbers. We used the true centroid criterion across all SENs to ensure consistent analysis.

Based on the aforementioned criteria, the block boundary changes for 2020 clearly exclude the area that includes the current-day apartment complex. The 2020 block (block 1015), that is included in Brutontown, is highlighted to the left. The updated block boundary excludes the area to the northeast. This area was included in 1990 using the same selection criteria. This boundary change likely explains much of the  difference in the population estimate for Brutontown  between 1990 and 2020.  

To clarify the Brutontown tabulations, convening residents is likely necessary to gain insight and clarity on what actually constitutes the neighborhood and which blocks should be excluded or included in analysis. The remaining twelve SENs were not significantly impacted by centroid definition or block boundary changes.

1. Chakraborty, J., & Armstrong, M. P. (1997). Exploring the use of buffer analysis for the identification of impacted areas in environmental equity assessment. Cartography and Geographic Information Systems, 24(3), 145-157.