Redistricting - Jimrtex, Alternate Process, Scoring System

[muon2]

So you are spreading the extra 0.3% in the non-compliant regions over all 10 districts. Links imply pairwise adjustments, but you use a globally determined correction. It works mathematically, but the graph isn't really needed to apply it in the way it would be if the adjustments were calculated on a pairwise basis. The only use for the graph is to determine if there are adjacent districts that are both in deficit or surplus implying that there is an extra shift to pass through population.

[jimrtex]

So you are spreading the extra 0.3% in the non-compliant regions over all 10 districts. Links imply pairwise adjustments, but you use a globally determined correction. It works mathematically, but the graph isn't really needed to apply it in the way it would be if the adjustments were calculated on a pairwise basis. The only use for the graph is to determine if there are adjacent districts that are both in deficit or surplus implying that there is an extra shift to pass through population.

In this case that is true, because all non-compliant regions were placed in a single area.  The effect of doing so will be to better equalize equality among all districts.  Hopefully this will produce plans with pretty good final equality, but without as much splitting up of counties.

Letting regions be closer to 0.5% deviation should provide greater flexibility.  I don't think you can really target to be just inside the limits, except in the northern districts.  But at the same time, it should produce smaller adjustments from fully county based plans.  Your scoring system may force some districts to be very close to equality, and then relatively large adjustments elsewhere.  This may result in overly constraining palns.

I 'm not sure I understand what you mean by "in the way it would be if the adjustments were calculated on a pairwise basis."   If I do understand what you are saying, I think my response is that is not necessarily a desirable alternative process.

Drawing a graph helps me understand what is happening, and I think it can help lead to better plans.  It is a useful visualization tool for me.

[jimrtex]

This is Muon's next plan from 2013.  It eliminates the double shift, but ends up with a greater total shift, likely because Region 2 now has a fairly large deficit.   The erosity is somewhat less.

[muon2]

So you are spreading the extra 0.3% in the non-compliant regions over all 10 districts. Links imply pairwise adjustments, but you use a globally determined correction. It works mathematically, but the graph isn't really needed to apply it in the way it would be if the adjustments were calculated on a pairwise basis. The only use for the graph is to determine if there are adjacent districts that are both in deficit or surplus implying that there is an extra shift to pass through population.

In this case that is true, because all non-compliant regions were placed in a single area.  The effect of doing so will be to better equalize equality among all districts.  Hopefully this will produce plans with pretty good final equality, but without as much splitting up of counties.

This is where I'm confused again. It looks like you are suggesting that that the shifts are actually what will be put in the plan. I thought the shifts were just a scoring tool for submitted plans. If they are just for scoring, then the actual shifts will be different from the calculated shifts. If the system imposes certain shifts between regions, we are back to the problem of scoring actual plans for the state that are submitted.

A scoring system has to be able to evaluate a plan for all the districts in the state. For instance it should be able to evaluate the one enacted into law. Otherwise it is a process to produce a plan, not a scoring system to evaluate plans.

[jimrtex]

So you are spreading the extra 0.3% in the non-compliant regions over all 10 districts. Links imply pairwise adjustments, but you use a globally determined correction. It works mathematically, but the graph isn't really needed to apply it in the way it would be if the adjustments were calculated on a pairwise basis. The only use for the graph is to determine if there are adjacent districts that are both in deficit or surplus implying that there is an extra shift to pass through population.

In this case that is true, because all non-compliant regions were placed in a single area.  The effect of doing so will be to better equalize equality among all districts.  Hopefully this will produce plans with pretty good final equality, but without as much splitting up of counties.

This is where I'm confused again. It looks like you are suggesting that that the shifts are actually what will be put in the plan. I thought the shifts were just a scoring tool for submitted plans. If they are just for scoring, then the actual shifts will be different from the calculated shifts. If the system imposes certain shifts between regions, we are back to the problem of scoring actual plans for the state that are submitted.

A scoring system has to be able to evaluate a plan for all the districts in the state. For instance it should be able to evaluate the one enacted into law. Otherwise it is a process to produce a plan, not a scoring system to evaluate plans.

The shifts are what will be placed in a Stage 1 Plan, and is what they will be evaluated on.  The actual placement of the adjustment will come in Stage 2.  The shifts identified in Stage 1 will be used as targets for Stage 2 - and can be used to choose among alternatives.

For example, your plan would set a shift of 5,822 from Region 8 (Detroit) to Region (7) Ann Arbor.  The actual transfer could come from Livingston, Oakland, or Wayne counties in to Washtenaw County.

When I scored the Torie and train plans, I removed their county fragments, but used them to identify the placement of inter-regional shifts.  The size of the inter-regional shifts was calculated on the whole-county region populations.  They will be similar to, but not identical to the size of the county fragments in the plan.

I don't see why the scoring system has to be able score the final plan enacted into law.

[jimrtex]

This is another Muon2 alternative from 2013.

[muon2]

So you are spreading the extra 0.3% in the non-compliant regions over all 10 districts. Links imply pairwise adjustments, but you use a globally determined correction. It works mathematically, but the graph isn't really needed to apply it in the way it would be if the adjustments were calculated on a pairwise basis. The only use for the graph is to determine if there are adjacent districts that are both in deficit or surplus implying that there is an extra shift to pass through population.

In this case that is true, because all non-compliant regions were placed in a single area.  The effect of doing so will be to better equalize equality among all districts.  Hopefully this will produce plans with pretty good final equality, but without as much splitting up of counties.

This is where I'm confused again. It looks like you are suggesting that that the shifts are actually what will be put in the plan. I thought the shifts were just a scoring tool for submitted plans. If they are just for scoring, then the actual shifts will be different from the calculated shifts. If the system imposes certain shifts between regions, we are back to the problem of scoring actual plans for the state that are submitted.

A scoring system has to be able to evaluate a plan for all the districts in the state. For instance it should be able to evaluate the one enacted into law. Otherwise it is a process to produce a plan, not a scoring system to evaluate plans.

The shifts are what will be placed in a Stage 1 Plan, and is what they will be evaluated on.  The actual placement of the adjustment will come in Stage 2.  The shifts identified in Stage 1 will be used as targets for Stage 2 - and can be used to choose among alternatives.

For example, your plan would set a shift of 5,822 from Region 8 (Detroit) to Region (7) Ann Arbor.  The actual transfer could come from Livingston, Oakland, or Wayne counties in to Washtenaw County.

When I scored the Torie and train plans, I removed their county fragments, but used them to identify the placement of inter-regional shifts.  The size of the inter-regional shifts was calculated on the whole-county region populations.  They will be similar to, but not identical to the size of the county fragments in the plan.

I don't see why the scoring system has to be able score the final plan enacted into law.

A scoring system should be able to take any submitted plan and either evaluate it with a score or scores or reject it as failing a specific rule of construction. A standard point of reference is to compare submitted plans to the one enacted into law.

[jimrtex]

A scoring system should be able to take any submitted plan and either evaluate it with a score or scores or reject it as failing a specific rule of construction. A standard point of reference is to compare submitted plans to the one enacted into law.

Here is the current legislative plan, as I have normalized it.



The current plan puts the Detroit UCC into 7 districts, permitting it to fan out into the Thumb, Washtenaw County, and to Lansing.  The Lansing UCC is split among 3 districts, with each of the three counties in a different district.

I first identified the 6 districts that had the largest share of the Detroit UCC, which left out MI-9 which includes Ingham, Livingston, and a big chunk of Oakland.  The remaining 6 districts made up Region 8 (Detroit).  I then assigned MI-9 to Region 4 (Lansinge) because it included the largest share of the UCC.  Region 3 (Grand Rapids) was the two district that covered the UCC.

Counties that were split were assigned to the region that contained the largest share of the county populations.

Next I reassigned the counties that had been chopped out of the UCCs to the region for the UCC.  This meant assigning Eaton and Clinton to Region 4.  Placing Clinton in Region 4 cut off Shiawassee, so it was placed in Region 4 as well.  Finally Livingston was assigned to Region 8.

The topmost population figures reflect the populations of Regions 2, 4, 7, and 8 with these reassignments.

The shifts indicated in red for Clinton and Shiawassee, lime for Livingston, and brown for Eaton reflect the whole county UCC chops in the legislative plan.  The second set of population figures represent these UCC chops restored to the legislative-drawn districts.   From the legislature's perspective  these were whole counties.  From our perspective they were massive UCC chops.

I then calculated the shifts to equalize population.   I ignored the small chop in Allegan County because it was not necessary to reach our population equality standards.  I also eliminated the Montcalm and Mason splits, in place of a split of Iosco.  A reasonable argument could be made that I was making the legislative plan a bit bette and it really should have scored at 127% rather than 125%.

[jimrtex]

This is the a chart of the scores with Muon2 and Jimrtex plans from 2013 added.



Blue squares are Torie's plans as evaluated by my scoring system.  They have a high level of inequality because they were not intended to be scored by my system.

Orange squares are Torie's plans with modifications made by me.

Red squares are Train's plans as evaluated by scoring system.  They also were not intended to be scored by my system.

Green squares are Train's plans as adjusted by me.

Green diamond are Jimrtex's plans from 2013.

Blue circles are Muon2's plans from 2013.

This is a chart with Torie's and Train's plans (as evaluated by my scoring system) removed, and the X-scale expanded,



The best plans are:

Muon2 2013B (3.0%, 1186 mi)
Muon2 2013C (3.3%, 1121 mi)
jimrtex 2013D (3.9%, 1035 mi)
jimrtex 2013E (4.8%, 1012 mi)
((torie 2013E))  (9.2%, 895 mil)