8.16.

Political Redistricting
Based on the new census information, it is time to redraw the boundaries of the political
districts in the state of Idazona. Each district will have one representative in the next
Congress, and Idazona has been allocated four representatives based on its share of the
national population. The state is made up of nine counties, with populations (in thousands)
shown in the table. (See the state map below.)

Figure 1: Caption

The main requirement in the formation of districts is that they produce equal populations,
or as close to equal as possible. Furthermore, the districts must be composed of adjacent
counties without splitting any county between two or more districts. Officials in Idazona
interpret the requirement to mean that if a district is created from two counties, then those
two counties must share a border. Furthermore, if a district is created from three counties,
then at least one of the counties must be adjacent to the other two. No district is permitted
to have exactly one county or more than three counties.
Mathematically, officials are seeking a districting plan for which the maximum deviation
between a district population and the average district population will be as small as possible.
What assignment of counties to districts will satisfy the desired conditions?

1
Solution
The problem parameters are the population in each county.
Let pi , for i ∈ {1, 2, 3, . . . , 8}, represent the population in county i.
The decision variables determine which counties are in which districts.
Let xi,j be a binary decision variable such that
(
1, if county i is in district j
xi,j = for i ∈ {1, . . . , 8}, j ∈ {1, . . . , 4}
0, otherwise

Each xi,j ∈ {0, 1}.

The objective is to minimize the maximum deviation between a district’s population and
the average district population.
The population of district j is given by:
8
X
pi xi,j
i=1

The average district population is: P8

i=1 pi
4
The objective function is:
8 P8
X
i=1 pi
min max pi xi,j −
j∈{1,...,4}
i=1
4

Each district must have exactly two counties:

8
X
xi,j = 2 ∀j ∈ {1, . . . , 4}
i=1

Counties in the same district must be adjacent. Let l1 and l2 be two counties that are not
adjacent:
xl1 ,j + xl2 ,j ≤ 1 ∀j ∈ {1, . . . , 4}, ∀(l1 , l2 ) not adjacent

Overall Model
8 P8
X
i=1 pi
min max pi xi,j −
j∈{1,...,4}
i=1
4

2
Subject to:
8
X
xi,j = 2 ∀j ∈ {1, . . . , 4}
i=1
xl1 ,j + xl2 ,j ≤ 1 ∀j ∈ {1, . . . , 4}, ∀(l1 , l2 ) not adjacent
xi,j ∈ {0, 1} ∀i ∈ {1, . . . , 8}, ∀j ∈ {1, . . . , 4}

Linearization
From the above, the objective function is not linear, so we will perform linearization. This
will build on top of the previous solution.
We introduce two new decision variables to handle the deviations:

dj for j ∈ {1, . . . , 4} (absolute deviation of district j)

y (maximum deviation across all districts)

We also define auxiliary variables:

uj , vj ≥ 0 for j ∈ {1, . . . , 4}

We want to minimize the maximum deviation:

min y

To linearize the absolute deviation, we use:

8 P8 !
i=1 pi
X
pi xi,j − + uj − vj = 0 ∀j ∈ {1, . . . , 4}
i=1
4

where uj , vj ≥ 0.
This constraint enforces that uj or vj captures the deviation from the average population.
If the expression inside the parentheses is positive, then uj will take its value and vj = 0. If
it is negative, then vj will take the positive value and uj = 0. Therefore:

dj = uj + vj

Next, to set y as the maximum of the deviations:

3
 y ≥ uj + vj ∀j ∈ {1, . . . , 4}

Since we are minimizing y, and y is constrained to be at least as large as each dj , it will be

equal to the maximum of the dj values at optimality.

min y
8
X
s.t. xi,j = 2 ∀j ∈ {1, . . . , 4}
i=1
xl1 ,j + xl2 ,j ≤ 1 ∀j ∈ {1, . . . , 4}, ∀(l1 , l2 ) not adjacent
y ≥ uj + vj ∀j ∈ {1, . . . , 4} (1)
8
!
P 8
X pi
pi xi,j − i=1 + uj − vj = 0 ∀j ∈ {1, . . . , 4}
i=1
4
uj ≥ 0, vj ≥ 0 ∀j ∈ {1, . . . , 4}
xi,j ∈ {0, 1} ∀i ∈ {1, . . . , 8}, j ∈ {1, . . . , 4}