Iowa Consolidation to 4 Districts

[jimrtex]

I have a novice question for JimTrex or anyone willing to answer. If another state was to take up Iowa's method, but a single county had significantly more population that a single Congressional district -- let's say Illinois' Cook County for simplicity's sake -- what would happen?

It would be declared unconstitutional.  Only states that have enough counties that none are that large are allowed to use this method.

The IA constitution requires whole counties, but the IA code requires the legislature to justify variances in excess of 1%. This is similar to hat happened in the OH Senate in 2011, where the legislature had to choose between conflicting laws. I assume they would divide the large county with as few chops as possible.

If not too off topic, could you elaborate on the conflicting laws that went up against one another for redistricting the Ohio Senate in 2011?

Ohio requires each senate district to be composed of three house districts. It also requires that counties be minimally divided among senate districts.

So if you have a county with 8 house districts. You would have two senate districts entirely in the county comprised of 3 house districts each, and another senate district comprised of two house districts in the county plus another house district. Simple.

But imagine you had three counties.

A(2) : B(8) : C(2)

6 of the House districts in B can form two senate districts. But what single district in A or C can be added to it? Even if A and C were adjacent, you would have to split one of those counties.

S1: B1, B2, B3
S2: B4, B5, B6
S3: B7, B8, A1
S4: A2, C1, C2

You have complied with the constitution with respect to B and C, but have divided A.

S1: B1, B2, B3
S2: B4, B5, B6
S3: B7, A1, A2
S4: B8, C1, C2

This version keeps A and C whole, but divides B.

My recollection is that the 2010 problem involved Cuyahoga County. Because the neighboring counties (other than Geauga) are large and have more than one house district, and there was no way to solve the problem. Northeast Ohio is constrained because you can't go into Lake Erie or Pennsylvania. Other areas of the state are less densely populated and you can go in more directions. The new apportionment standards for Ohio, at least say that if you have to violate constraints, you should do the fewest possible.

[jimrtex]

The Give and Take Process is designed to balance district populations by gradually adjusting populations. While it splits counties, this is done under control of the voters, and presumably less susceptible to gerrymandering.

In this application, after each equalization pass, district boundaries will be reset to whole counties based on the district with the majority of the population.



Initially Southeast has large deficit (0.194), Central has a large surplus (0.136), Northeast had a moderate surplus (0.061), and West has a tiny deficit (0.002).

The divisor for the first round is 10, and Southeast takes (0.194/10) = (0.019) from any of its three more populous neighbors. It chooses Washington. While we could track townships, I just kept track of the total.

Central gives (0.136/10) = (0.014) to any of its three smaller neighbors. It prioritizes the three counties not in the Des Moines CSA as follows: (1) Marshall, which is not part of the CIRTA RCOG, and not adjacent to Polk because of the boundary offset; (2) Marion because it is more rural, and the Des Moines River is somewhat disruptive ti travel Pella; and (3) Jasper because it is a straight shot to Newton down I-80, and therefore in easy commuting range. So the first transfer is Marshall to Northeast.

Northeast has lost 0.019 and gained 0.014 giving it a surplus of 0.056, and must give 0.006 to a smaller neighbor. It doesn't want to lose territory to the west, as some folks are agitating to expand westward and include Cerro Gordo (Mason City). So it goes along with the transfer of Washington.

West needs to take a tiny amount from a larger neighbor. It chooses the more rural Mitchell, rather than Boone or Marion.

The second round with a divisor of 9 repeats the process, except Southeast has completed the transfer of Washington, and begins taking of Johnson.

By the fourth round the transfer of Marshall to Northeast is complete, and the transfer of Marion begins. At this point populations are Southeast 0.901, Central 1.081, Northeast 1.019, and West 0.999.

By the sixth round, Northeast is pulled below a quota, and must take from a larger neighbor. It choose to take from West, quickly reversing the transfer of Mitchell, and starts to take from Worth. West is forced to take from Central, since it can't take back from Northeast and begins to nibble from Boone.

By the seventh round, the transfer of Marion has been completed, and the transfer of Jasper begins. At this point the populations are Central 1.038, Southeast 0.977, Northeast 0.989; and West 0.995. Central now has the largest deviation and goes first, continuing the transfer of Jasper.

In the 9th round, the order of West and Northeast reverses, and West begin to take back in Worth, while Northeast is forced to take from Central, taking a bit more from Jasper.

Before the 10th Round the populations are Central 1.011, Southeast 0.992, West 0.998, and Northeast 0.999. The divisor for the round is 1.

Central continues to transfer Jasper. This gives Northeast a population of 1.010, and the largest deviation. Northeast gives to Southeast. This gives Southeast a population of 1.002 and West a population of 0.998, and the two are equalized with a tiny transfer of Wayne.

All districts have equal population. But we move counties into districts with which they have the largest population in.

Washington and Johnson are transferred from Northeast to Southeast.
Marshall and Jasper from Central to Northeast.
Marion from Central to Southeast.
Worth from West to Northeast.

[jimrtex]

After completion of the first 10 rounds of Give and Take, the total deviation has decreased from 0.392 to 0.144. Because all of Johnson was transferred Southeast now has the surplus and Northeast the deficit.

Populations in order of absolute deviation: Southeast 1.072, Northeast 0.942, West 0.989, Central 0.997.

It is possible to construct a district in eastern Iowa including Cedar Rapids, Dubuque, Davenport and Iowa City, but without major swaps in order to encompass Cedar Rapids, this is out of range. So the decision for Southeast was whether to get rid of Dubuque in order to make a more compact district, return Marion to Central (or Northeast) or move rural counties to West.

The southern counties want the counties to the west to serve as a counterbalance to the cities along the river. Johnson (Iowa City) holds the balance of power, and ultimately votes to Give Dubuque to Northeast making the district more centered on Iowa City.

Northeast has a deficit and must Take from any of its three larger neighbors. There is little sentiment to take Story other than from ISU alumni, and there are many UNI and Iowa alumni to reject this. So the choices are to take Dubuque or expand to the west. Ultimately the decision is made to take Cerro Gordo (Mason City). "That will put us right close to a quota:.

West has a deficit and must take from Central or Southeast. Some want to nibble at Boone, but the final decision is to take Wayne from Southeast. Central takes tiny bits of Marion from Southeast.

In the fourth round, Wayne is completely transferred to West, and they begin taking from Lucas.

At the start of the 5th round the populations are:

Southeast 1.035, West 0.983, Northeast 0.985, Central 0.998.

This has reversed the order of Northeast and West, so the West can now take from Northeast, and begins to take back Cerro Gordo. Northeast is force to take from Southeast or Central, and begins to take Dubuque.

This pattern continues through the 10th round.

At that point, West had taken back all of Cerro Gordo, and had just barely began to take from Worth. West had taken all of Wayne and a tad of Lucas, and keeps Wayne.

48.6% of Dubuque had been transferred to Northeast. Since this is not a majority, Dubuque stays with Southeast.

Some observers are wondering whether this system will work out. Others say, let's shift Wayne and see if it makes a difference.

[jimrtex]

At that point, West had taken back all of Cerro Gordo, and had just barely began to take from Worth. West had taken all of Wayne and a tad of Lucas, and keeps Wayne.

48.6% of Dubuque had been transferred to Northeast. Since this is not a majority, Dubuque stays with Southeast.

Some observers are wondering whether this system will work out. Others say, let's shift Wayne and see if it makes a difference.

After another round Lucas was transferred to West, but then it got stuck as it proved impossible to flip Dubuque to Northeast.

Adding another rule. that the largest county had to be switched, so Southeast instead shifted Marion to Northeast. But this seemed a little bit like 'jimrtex ex machina'. Iwould have to whisper in the ears of the voters to make it work.

[jimrtex]

In this versions, I set it up so that there was gradual flow between all adjacent regions from the more populous to the less populous, with the flow proportional to the difference. In the starting map. all districts are adjacent to each other. This would not be true in all states, but it is for this version of an Iowa map (because of few districts, and the central district).

After the equalization flows were identified whole counties were associated with the flows. In actual practice, the voters of the receiving district could choose the counties. It is fundamentally better that a district can choose which areas that they wish to add to their district, rather than which areas are to be removed.

Flows:

Central to Southeast 0.084: Marion and Jasper.
Northeast to Southeast 0.064: Washington, Poweshiek, Iowa
West to Southeast 0.048: Clark, Decatur, Ringgold, Union
Central to West 0.034: Boone
Central to Northeast 0.018: None
Northeast to West 0.016; Mitchell

This results in the following map.

[muon2]

Could you explain in a little more detail the workings of "flow proportional to the difference"?

[jimrtex]

Could you explain in a little more detail the workings of "flow proportional to the difference"?

Initially the populations were

C 1.135, NE 1.061, W 0.998, SE 0.806

Between two adjacent districts, A and B, with populations P_A and P_B, the flow F_AB is (P_A - P_B)/D

For the first round I used a divisor of 10, so the flow from C to SE was (1.135 - 0.806)/10 = 0.016. Flows between other adjacent districts were calculated in a similar fashion, and the populations were updated,

C 1.108, NE 1.049, W 0.999, SE 0.844

And the process repeated with a divisor of 9.

There may be physical analogues (current flow between capacitors? fluid flow between tanks?).

I suppose that a constant divisor could be used, but then convergence to equilibrium would take infinitely long. I also had envisioned a smaller set of discrete steps since the voters would intervene. For that initial flow of 0.016 from Central to Southeast, Southeast could choose between Marion, Jasper, and Warren. They would of course have chosen Marion because it fits much better with the district.

Once they breached the county line, all transfers between Central and Southeast would be in Marion, until all of Marion was in one district or the other. There could be much smaller discrete steps, with the voters only queried when a county had completely changed hands. So the less populous district in each pair would choose a target county. The computer would then run in small steps until a county had been completely transferred, and then the voters would choose a new target.

A problem with that scheme might be that Southeast could have chosen Johnson after Washington had been transferred. After equilibrium was reached Johnson would have been less than 50% transferred. This happened between Central and Northeast, where Northeast chose Marshall. Perhaps there could be look-ahead, and Southeast could be told that they could not choose Johnson.

There is also the possibility that a pair of districts coulld be cut off from each other, or a new adjacency created. Using small discrete steps could handle that situation. Even though envisioned as continuous system, it could reconfigure the model at a discontinuity.

PS there is another step coming, to reach better equality.

[jimrtex]

We can make variants, which we can rate according two criteria; Persons shifted to a new district, and equality as measured by standard deviation. Few persons shifted, and smaller standard deviation are better;



This variant has shifted 0.109 quotas (Marshall, Boone, Floyd, and Ringgold) and has a standard deviation of 0.468%. The unchanged version shifts 0.000 quotas and has a standard deviation of 1.456%



This second variant shifts Howard, resulting in a shift of 0.121 quotas, but reduces standard deviation to 0.405%.



This third variant shifts Marshall to Northeast, but rather than adding Boone to Central, adds Adair, Union, and Clarke to Central, which gets it closer to the quota. Jones is moved to Southeast, which gets Northeast close to the quota. Ringgold is the final adjustment. This map has a shift of 0.118 quotas, but has reduced the standard deviation to 0.047%.

Which Plan is Best?

Why don't we ask the voters? We have them rank the plans. But we don't have them rate the whole map. What someone in Sioux City thinks about the Davenport district is irrelevant. Instead we have the voters in each county rank the districts that they would be placed in. That ranking is then applied to the overall plan. If a district is the same in more than plan, then those counties would rank all such plans equally.

The best plan is the Condorcet winner. If there is a cycle we could break it by eliminating the plan that has the greatest standard deviation.

[muon2]

Initially the populations were

C 1.135, NE 1.061, W 0.998, SE 0.806

Between two adjacent districts, A and B, with populations P_A and P_B, the flow F_AB is (P_A - P_B)/D

For the first round I used a divisor of 10, so the flow from C to SE was (1.135 - 0.806)/10 = 0.016. Flows between other adjacent districts were calculated in a similar fashion, and the populations were updated,

C 1.108, NE 1.049, W 0.999, SE 0.844

And the process repeated with a divisor of 9.

There may be physical analogues (current flow between capacitors? fluid flow between tanks?).

I suppose that a constant divisor could be used, but then convergence to equilibrium would take infinitely long.

The constant in electrical or fluid flow depends on the size of the pipe. If you want to consider physical flow as a model the formula might be F = k E (delta P), where k is a constant for all rounds, E is proportional to the boundary length or erosity between the two districts, and delta P is the population difference. k should be large enough that overshoots should be possible, since oscillations can happen in underdamped systems. Erosity could also be applied to the relative county population, so that the counties most contributing to erosity are most likely to move.

[Torie]

Jimrtex's map is an artistic failure. Thank you.

[President Punxsutawney Phil]

How does this map compare?

[Torie]

How does this map compare?

That is quite nice, but just how does the Des Moines based CD add three more counties to what Jimrtex had without having too many people?

[America Needs a 13-6 Progressive SCOTUS]

How does this map compare?

That is quite nice, but just how does the Des Moines based CD add three more counties to what Jimrtex had without having too many people?

probably using different population totals.

[President Punxsutawney Phil]

I used 2010 census data for that one...in any event, 2020 will require some adjustments.

[jimrtex]

Initially the populations were

C 1.135, NE 1.061, W 0.998, SE 0.806

Between two adjacent districts, A and B, with populations P_A and P_B, the flow F_AB is (P_A - P_B)/D

For the first round I used a divisor of 10, so the flow from C to SE was (1.135 - 0.806)/10 = 0.016. Flows between other adjacent districts were calculated in a similar fashion, and the populations were updated,

C 1.108, NE 1.049, W 0.999, SE 0.844

And the process repeated with a divisor of 9.

There may be physical analogues (current flow between capacitors? fluid flow between tanks?).

I suppose that a constant divisor could be used, but then convergence to equilibrium would take infinitely long.

The constant in electrical or fluid flow depends on the size of the pipe. If you want to consider physical flow as a model the formula might be F = k E (delta P), where k is a constant for all rounds, E is proportional to the boundary length or erosity between the two districts, and delta P is the population difference. k should be large enough that overshoots should be possible, since oscillations can happen in underdamped systems. Erosity could also be applied to the relative county population, so that the counties most contributing to erosity are most likely to move.

I just thought of another possibility for conductance; voter preference. There are 13 counties adjacent to Southeast in districts with greater population. Voters could rank them, and then the lowest choices would be eliminated in turn, until one county remained for each adjacent district. Conductance would be weighted based on the preference ratio among the three counties.

Or there could also be voting in the other districts to determine which counties were shed. So the voters

Southeast choose between 13 counties to shift to SE.
Northeast choose between 12 counties to shift from NE, and 2 to shift to NE
West choose between 10 counties to shift to W, or 2 to shift from W.
Central between 10 counties to shift to other districts, including the destination district for Story, Jasper, and Warren.

Preferences would be shifted until one county was chosen for each pair of adjacent districts, and it would have a weight based on intensity of support.

I'm concerned about overshoot since that might cause discontinuities when counties shifted.

[jimrtex]

How does this map compare?

This uses the 2020 county projections.



73 of 99 Iowa's counties are losing population.

The only counties above 10% growth are Polk, Dallas, Warren (Des Moines metro), Story (Ames), Johnson (Iowa City), and Jefferson (Fairfield). Linn (Cedar Rapids), Dubuque (Dubuque), and Scott (Davenport) are somehat below. Black Hawk (Waterloo), Woodbury (Sioux City), and Pottawattamie (Council Bluffs) are gaining anemically. Dallas is close to 50% gain.

If you take your map and shift Marshall and Jasper to the east, and Greene to the west, and then move some counties from Northeast and Southeast to the west, you would have a 2020 map. This could have been an approach if your plan had been adopted after 2010, and preserving cores of existing districts was important.