Hands-Off Redistricting Strategies

“Hands-Off” has a “Strategy”?

Having argued the case that Congressional districts should be drawn in a hands-off way, it may seem contradictory to write about “strategies” for redistricting.  But the strategies I am thinking about today don’t have to do with political strategies for political ends (e.g. strategies for how to dilute minority votes, or how to make sure the incumbent congressperson wins), but instead pertain to technical strategies for how to best solve the Five Rules for Hands-Off Redistricting. One has to think about this for two reasons:

  1. When you know the best strategy for solving the problem, you will be a long way toward understanding if a deterministic or probabilistic approach is required, and 
  2. The best technical strategy will give insights into the overall character of congressional districts that will result.

When I think about OPRA and how it would be best satisfied, it seems intuitively that the optimal solution would involve these two types of districts:

  1. Some number of densely-populated, compact districts that have many short boundaries with other densely-populated districts, combined with some boundaries with sparsely-populated districts
  2. A few sparsely-populated rural districts with long boundary lines between them.

It is tempting to just go ahead and call these urban and rural districts, and it would seem that the OPRA solutions game would go like this:

1. Define the optimal number of compact/closely-packed urban districts that contain the target number of residents with the use of very short Fences.

2. Once the optimal number of urban districts has been created, it is left to divide up the remaining area of the state with equally-populated rural districts.

It seems to me that the urban districts would be segmented by something akin to a compactness or close-packing algorithm– MLE in San Francisco has suggested as much to me. Once the urban districts are defined, though, it seems like there will have to be a “dash to the borders” wherein the game will be to find the shortest arrangement of (fairly long) fences that do the remaining required segmenting.

An illustration using California

It is probably best to illustrate what I mean by this. Based on current census data, California has about 38m residents and is apportioned 53 Congressional seats, meaning that the target population for each of the 53 districts is about 717,000 people.

I guess that any SMSA with over 400,000 people in it would make one or more “compact” districts, and I find 15 such SMSA’s in the state, ranging from a high of Los Angeles-Long Beach-Santa Ana with almost 13m people (worth 18 districts) and a low of Vallejo-Fairfield which at best could be the center of only one district.  If I crudely plot these urban districts on a map (not worrying about how to divide large SMSA’s like Los Angeles for now) I get this:

CA urban blobs

This represents 46 of the 53 districts available to us, and presumably also corrals a total of about 33m people at 717,000 people/district. So the remaining 5m people and their 7 remaining districts have to come out of the region that is not contained in blue lines. Now the game changes: where before you were working with cellular-like urban districts, now you want to equivalently segment very large areas with the least amount of fence. This was the “dash to the borders” task I mentioned above, it also seems akin to lightning finding the path of least resistance to state or urban cell borders.  Hypothetical lightning-strike fences are now added to the map in red:

CA all blobs

They are not straight, remember, because they have to follow existing Census Tract boundaries.  And of course keep in mind that this whole exercise is baloney, because I don’t even know the populations of these rural areas. (And also I regret using red and blue because that isn’t the point!) But IF this is the best way to approach an OPRA solution, I think it is revealing just as a thought experiment for a couple of reasons.

First, I think this is a promising result.  You really will vote with your neighbors if you are in the blue areas, and if you are in the rural areas you are (by rural standards) voting with people facing similar issues as you.  I’m not going to say that when it comes to finding the best blue fence lines that some cities (or school districts, or fire districts, etc.) aren’t going to get segmented, but they are now and at least this approach guarantees more compactness of residents within districts. A random dose of representative diversity in each district is a good thing.

Second, this paradigm of cellular urban districts and “lightning-strike-segmented” rural districts may frame the discussion about whether OPRA is Deterministic or Probabilistic, and if it turns out it is probabilistic, it may establish a framework of heuristics about how to approach looking for a solution.

Still looking for help with the math! If you have any thoughts on this approach, or know anyone who might find it interesting, please let me know!

About noahkennedy

non-fiction author
This entry was posted in Boundary mathematics, OPRA and tagged , , . Bookmark the permalink.

2 Responses to Hands-Off Redistricting Strategies

  1. Pingback: Brian Olson’s approach to Hands-Off Redistricting | Fair and Sane Redistricting

  2. Pingback: A GIS analysis of Gerrymandering compared to OPRA | Fair and Sane Redistricting

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s