Tuesday, August 21, 2018

​Gerrymandering - There Are Simple Solutions To This Problem

Published in HuffPost
September 21, 2017 - 04:40 pm ET
By Rich Weissman, Palm Springs, California (www.richweissman.com)

Gerrymandering is a pernicious method allowing one political party control by manipulating electoral districts, producing a body politic of heightened polarization. It has become such an insidious way of destroying voter confidence and fairness that it has recently gone before the U.S. Supreme Court for a ruling. During the SCOTUS debate, Ruth Bader Ginsburg noted to Neil Gorsuch when he questioned the issue relative to the role of the Supreme Court, she replied, “Where did ‘one person, one vote’ come from?” The Equal Protection Clause of the Constitution. Clearly, gerrymandering is an issue that SCOTUS must address, as the practice diminishes this clause as a fundamental foundation of democracy.

I asked myself how might we envision creating electoral districts blind to political outcomes? To be sure, there isn’t one answer, but there are solutions that are better than today’s method. I challenged myself to think of ways that would improve upon our current partisan approach, and devised one that uses basic geometry and algebra, along with statistical modeling, for its solution. I devised a simple one, perhaps worthy of adding to the mix of solutions on the table. It serves as an example of how we can think about the gerrymandering menace, and provide ideas for district mapping that are population and geography concentrated, and not based on other factors.

My background is in statistical analysis and data optimization, so the idea is based on a very simple use of optimization modeling. The concept is to determine the optimal districting within each state based on the solution that minimizes the shape of each district into its most efficient and concentrated and contiguous configuration, independent of all other variables other than population counts and geography. It’s a simple two-variable model, which would eliminate bias from any other variables (e.g. voting patterns, demographics), strictly mathematically driven without an eye to voting outcomes.

For each state, the model would use the population counts at the discreet block level (the lowest level of census data – not block group nor census tract levels) as per the U.S. Census for all blocks within the entire state. By entering the number of districts needed for the state (let’s call this number “X” for any given state), it would use a simple algorithm to determine the optimal configuration of all possible combinations of contiguous blocks that form “X” districts that minimize the sum of the perimeters of all “X’ districts and that produce districts of relatively the same numbers of population (total population of the state divided by “X”). Of all possible permutations (outcomes), the one permutation which produces the lowest value for the sum of each of the perimeters combined for all “X” districts, each with about the same population counts in a continuous shape, would become the optimal districting map solution for that state. It forces the map to provide districts that are as simply and efficiently shaped as possible, based solely on population counts. The geometric area of the districts would not be the driver nor enter the equation, as some parts of each state are sparsely populated and large in area. And, gerrymandering can occur in districts where the area of the districts are small, but the perimeters are large because the districts are designed with convoluted shapes. 

However, using this idea of perimeter minimization, the issue of districts is not the area of the district, but its perimeter. Think of a square of 3 miles X 3 miles (a square is used in this example as it is an efficient 4-sided polygram – the circle or more circle-like polygrams, such as an octagon, are more efficient but not used in this example to keep the example simple). Its area is 9 square miles, and its perimeter is 12 miles long (the sum of the sides of the square). Now think of a long rectangle 1 mile by 9 miles (assuming its population count is comparable). Its area is also 9 square miles, but its perimeter is 20 miles long. The 3 miles X 3 miles square is more efficient as a district as its perimeter is less and it represents the most efficient way to cover 9 squares miles in this example. This is fundamentally how the model could work, where the perimeter of each district would be the driver. One wouldn’t wind up with perfect squares or other polygrams (or circles) necessarily, but with configurations that minimize perimeters in the most efficiently shaped districts possible, according to how the population distributes itself geographically by blocks.

This would be a relatively easy algorithm to create, program and run the data, and would eliminate gerrymandering. Of course, this, like other ideas, would need to be tested with real data, and all ideas would be compared to see how the different solutions look.

The point is that the idea noted above demonstrates that there are ways to configure electoral districts that eliminate gerrymandering and mitigate political and other bias, through the utilization of mathematical models based on data optimization. And, it’s time for our political system think in new ways, including utilizing database technologies for optimization modeling, to protect and uphold the “one person, one vote” principle.