What if the area you’re interested in doesn’t completely cover a whole census tract? Assuming a uniform distribution of people throughout the census tract, you can weight the demographics you take from each tract based on the area of intersection between the census tract and your shape of interest:
This won’t give you a completely accurate figure but it will be meaningful as a representation of the areas you’re trying to examine.
Here’s a visual example using the 17 7-Eleven stores in Manhattan (the black lines delineate census tracts):
The 100 meter buffer around the store at 177 Dyckman St includes about 10% of the area in census tract 36061029100, about 0.7% of the area in census tract 36061028700, and about 0.13% of the area in census tract 36061028500.
Applying these weights to the demographics from each census tract gives me the aggregate numbers for that area. After dividing by the total population, the 100 meter area around this 7-Eleven looks like this:
|median age male||31.88|
|median age female||35.77|
|percent white, nonhispanic||0.05|
|percent black, nonhispanic||0.04|
|percent asian, nonhispanic||0.01|
|percent owner occupied housing||0.03|
|percent renter occupied housing||0.97|
So what does it look like when we compare the 7-Elevens to Manhattan as a whole? A following post will reveal the hidden secrets of 7-Eleven locations around Manhattan.