Blind Partitioning
A Simple, Unbiased Algorithm for Delimiting Polling Division
Suppose we want to divide the Colombo Electoral District into 16 polling divisions. We could start with dividing the region into two sub divisions with roughly equal populations, each with 8 seats.
How precisely do we do the division? We could make the choice based on various criteria. For example, we could consider Colombo’s shape (longer on the East-West longitude, shorter on the North-South latitude) and divide on a North-South axis.
Alternatively, we could consider some other criteria like preventing Gerrymandering (see Measuring Gerrymandering).
Let’s pick the former, and we get the following:
(Note, we have respected Grama Niladari Division boundaries. Hence, the “bumpiness”)
We continue with the division process until get end-up with 16 polling divisions:
Note, this method also works for divisions which are not powers of 2. For example, if we wanted to divide Colombo into 20 divisions, we first divide into 2 and 4 divisions, subsequently divide the 4ths in a 3–2 ratio, divide the 3 into 2–1, and so on.
