Establishing base line data for the forthcoming redistricting is a multi-step procedure which can be effectively assisted with GIS.
In the map and table below I’ve calculated out the total population for each district, along with a preliminary data variable, in this case African-American population.
Unc. = uncontested election.
Austin Scott is Rep. gain in GA-08.
Unfortunately for analysis, congressional districts (CDs) do not necessarily follow convenient borders. For the most part they follow counties but can veer off at any time. This splits counties between different districts.
Nor can CDs be aggregated from smaller units such as tracts or block groups. Whoever drew the current districts in Georgia (and I’d like to meet them!) felt free to veer around cities and suburbs as they wished.
What I’ve done then in this first approximation is as follows.
I used Esri MapData10 provided with our license. This includes 2010 data for counties, census tracts etc. (Presumably this is calculated out by Esri from the census.)
1. Intersect counties with the CDs. This produces polygons for all counties completely in a CD and all parts of counties.
2. Calculate the area of each unit (square meters).
3. Calculate the ratio of each unit to its original county.
4. Multiply the results of step 3 against your desired variables such as total population, African-American population, etc.
What this does therefore is produce a population estimate for each unit according to the area. An assumption here (which is not always valid) is that population is distributed evenly across the county. Obviously in a largely rural county with a town in one corner this assumption is violated.
A solution to this is to use smaller units, for example tracts or even block groups rather than counties. This is a computationally intensive but viable solution. However, it would be nice to know how much more accuracy that method buys you. If it is less than 5% counties may be the best bet in practice.