Skew in state delegations

[muon2]

A fair map in TN is 6R-3D.

Actually not. TN is R+14, or in a 50-50 national race the state should split 64% R to 36% D. Studies have shown that for every point that a state swings from 50% the delegation swings by 2%, so that in a 50-50 national race the TN delegation would be expected to be 78% to 22%. With 9 seats that corresponds almost exactly to a 7R - 2D delegation.

[muon2]

If you want strict proportionality you oughta just opt for PR tim.

I don't need to adopt PR.
It's possible to get rough proportionality in many places (like TN) and I don't think it's all that terrible trying to achieve it, even if one keeps single-member districts.

Think about it this way. Assume a population of 100 voters in 10 counties, and turnout is 100%. Let's say the state is 60 R and 40 D and the breakdown by county is like this:
A: 8 R 2 D
B: 7 R 3 D
C: 7 R 3 D
D: 6 R 4 D
E: 6 R 4 D
F: 6 R 4 D
G: 6 R 4 D
H: 5 R 5 D
I: 5 R 5 D
J: 4 R 6 D
If I don't worry about connections but group these into 5 districts, the best I can do is 1 Dem district, at that's only by combining J with H or I. 60% vote becomes an 80% delegation. It's likely that a geographic split would result in a 10-10 district at best for the Dems - with the parties reversed this is like the problem the Pubs have in MA.

Suppose one of the counties is a heavily Dem city, and the counties are like this.
A: 7 R 3 D
B: 7 R 3 D
C: 7 R 3 D
D: 7 R 3 D
E: 7 R 3 D
F: 6 R 4 D
G: 6 R 4 D
H: 6 R 4 D
I: 5 R 5 D
J: 2 R 8 D
Here the Dems are guaranteed 1 district, by matching J with any other county, but that's all they can get. Again a 60% vote results in an 80% delegation. This is like the situation in TN if there were only 5 districts and J was Shelby county.

Partitioning a state into single member districts will naturally enhance the advantage of the majority party in the legislature without any gerrymandering. Beyond these simple examples there are a number of academic studies that come to the same conclusion.

[muon2]

A fair map in TN is 6R-3D.

Actually not. TN is R+14, or in a 50-50 national race the state should split 64% R to 36% D. Studies have shown that for every point that a state swings from 50% the delegation swings by 2%, so that in a 50-50 national race the TN delegation would be expected to be 78% to 22%. With 9 seats that corresponds almost exactly to a 7R - 2D delegation.

I've seen you say this a lot, can you link me to some of these studies

One recent paper that cites a number of the studies is Gerrymandering or geography? How Democrats won the popular vote but lost the Congress in 2012, N. Goedert (Research and Politics 2014). The particular 2 to 1 effect is well documented in studies, some going back to Tufte in 1973. My metrics of skew and polarization mirror Tufte's goals of plans being unbiased and responsive.

[muon2]

A fair map in TN is 6R-3D.

Actually not. TN is R+14, or in a 50-50 national race the state should split 64% R to 36% D. Studies have shown that for every point that a state swings from 50% the delegation swings by 2%, so that in a 50-50 national race the TN delegation would be expected to be 78% to 22%. With 9 seats that corresponds almost exactly to a 7R - 2D delegation.

I've seen you say this a lot, can you link me to some of these studies

One recent paper that cites a number of the studies is Gerrymandering or geography? How Democrats won the popular vote but lost the Congress in 2012, N. Goedert (Research and Politics 2014). The particular 2 to 1 effect is well documented in studies, some going back to Tufte in 1973. My metrics of skew and polarization mirror Tufte's goals of plans being unbiased and responsive.

Can this be explained by statistical modeling, and do the underlying assumptions of the model hold true for all places and all times?

Some of the studies have done statistical simulations, but I don't know how many of the assumptions have been independently tested.

[muon2]

My system works best when applied on nationwide scale, so you can compensate when geography makes it impossible to have proportionality and reasonable districts at the same time.
A good example is Massachusetts. Simple proportionality is impossible in MA while also maintaining reasonable districts.
The solution would be making VA or PA have a more pro-GOP map to compensate.
Et cetera...

That's distinctly unfair to the people of the states like PA or VA. Why should they have districts due to the residency and voting patterns of MA? Each state is sovereign and should have a fair map that reflects its own population.

[muon2]

My system works best when applied on nationwide scale, so you can compensate when geography makes it impossible to have proportionality and reasonable districts at the same time.
A good example is Massachusetts. Simple proportionality is impossible in MA while also maintaining reasonable districts.
The solution would be making VA or PA have a more pro-GOP map to compensate.
Et cetera...

That's distinctly unfair to the people of the states like PA or VA. Why should they have districts due to the residency and voting patterns of MA? Each state is sovereign and should have a fair map that reflects its own population.

Well, this was tailor-made for those cases I drew House district maps for with all 50 states in mind.
It wouldn't be workable to completely copy this method in real life.

But how much weight should one place on partisan fairness? Intentional partisan unfairness should be avoided, but that's different from intentional partisan fairness. IA is often cited as the gold standard for neutral redistricting at the congressional and state levels, but the mappers are barred from using any election data. AZ has a neutral commission, but faced all sorts of issues this cycle because they selected certain data sets in an attempt to create nominally fair and competitive districts.

The central issue I see is that voting patterns change with time. An attempt to be fair in 2011 may not result in the appearance of fairness in 2017. I demonstrated that with my remap of the WI legislative districts last year. The same plan that looked like a pro-Dem map after 2008 (when it would have been drawn) looked like a pro-Pub map after 2016 due to the shift in rural WI voters.

Checking for unnecessary partisan skew or polarization I think should be done. I don't draw my plans while looking at the PVI's, but I do check the partisan results afterwards. I know attempts to lock in fairness will often fail beyond the first election. They even could backfire depending on the shifts in the electorate.

[muon2]

My system works best when applied on nationwide scale, so you can compensate when geography makes it impossible to have proportionality and reasonable districts at the same time.
A good example is Massachusetts. Simple proportionality is impossible in MA while also maintaining reasonable districts.
The solution would be making VA or PA have a more pro-GOP map to compensate.
Et cetera...

That's distinctly unfair to the people of the states like PA or VA. Why should they have districts due to the residency and voting patterns of MA? Each state is sovereign and should have a fair map that reflects its own population.

Well, this was tailor-made for those cases I drew House district maps for with all 50 states in mind.
It wouldn't be workable to completely copy this method in real life.

But how much weight should one place on partisan fairness? Intentional partisan unfairness should be avoided, but that's different from intentional partisan fairness. IA is often cited as the gold standard for neutral redistricting at the congressional and state levels, but the mappers are barred from using any election data. AZ has a neutral commission, but faced all sorts of issues this cycle because they selected certain data sets in an attempt to create nominally fair and competitive districts.

The central issue I see is that voting patterns change with time. An attempt to be fair in 2011 may not result in the appearance of fairness in 2017. I demonstrated that with my remap of the WI legislative districts last year. The same plan that looked like a pro-Dem map after 2008 (when it would have been drawn) looked like a pro-Pub map after 2016 due to the shift in rural WI voters.

Checking for unnecessary partisan skew or polarization I think should be done. I don't draw my plans while looking at the PVI's, but I do check the partisan results afterwards. I know attempts to lock in fairness will often fail beyond the first election. They even could backfire depending on the shifts in the electorate.

I suppose one could alter the formula to inject swing districts into the equation, when that is called for. In the case of WI, for instance, one could have 3D, 3R, and 2 swing.

My skew and polarization formulas do treat swing districts differently.

[muon2]

My system works best when applied on nationwide scale, so you can compensate when geography makes it impossible to have proportionality and reasonable districts at the same time.
A good example is Massachusetts. Simple proportionality is impossible in MA while also maintaining reasonable districts.
The solution would be making VA or PA have a more pro-GOP map to compensate.
Et cetera...

That's distinctly unfair to the people of the states like PA or VA. Why should they have districts due to the residency and voting patterns of MA? Each state is sovereign and should have a fair map that reflects its own population.

Well, this was tailor-made for those cases I drew House district maps for with all 50 states in mind.
It wouldn't be workable to completely copy this method in real life.

But how much weight should one place on partisan fairness? Intentional partisan unfairness should be avoided, but that's different from intentional partisan fairness. IA is often cited as the gold standard for neutral redistricting at the congressional and state levels, but the mappers are barred from using any election data. AZ has a neutral commission, but faced all sorts of issues this cycle because they selected certain data sets in an attempt to create nominally fair and competitive districts.

The central issue I see is that voting patterns change with time. An attempt to be fair in 2011 may not result in the appearance of fairness in 2017. I demonstrated that with my remap of the WI legislative districts last year. The same plan that looked like a pro-Dem map after 2008 (when it would have been drawn) looked like a pro-Pub map after 2016 due to the shift in rural WI voters.

Checking for unnecessary partisan skew or polarization I think should be done. I don't draw my plans while looking at the PVI's, but I do check the partisan results afterwards. I know attempts to lock in fairness will often fail beyond the first election. They even could backfire depending on the shifts in the electorate.

I suppose one could alter the formula to inject swing districts into the equation, when that is called for. In the case of WI, for instance, one could have 3D, 3R, and 2 swing.

My skew and polarization formulas do treat swing districts differently.

How do you define a swing district under your system?

These next two items will deal with the political measures for a plan. Under Item 3, they will only be used to guide the commission when they consider final plans, and can't be used to eliminate plans prior to the end of the submission period. First I will provide some definitions.

Definition: PVI. The Partisan Voting Index is based on the method developed by the Cook Political Report and widely used in assessing the political tendencies of congressional districts. It compares the Democratic share of the two-party vote in a state or a district to the Democratic share of the national presidential vote, averaged over the last two presidential elections. Shares are multiplied by 100 to get a percent, and expressed as D+x when positive and R+(-x) when negative.

Definitions: A highly competitive district has a PVI of 0 or 1 (-0.014 to +0.014) and statistically such districts have an even chance of being one by either district. A competitive district has a PVI of 2 through 5 (-0.054 to -0.015 and +0.015 to 0.054) and statistically such districts have an 3 out of 4 chance of being held by the favored party. An uncompetitive district has a PVI of 6 or greater for either party and has better than a 9 in 10 chance of being held by the favored party. This is based on congressional results during the preceding decade.

Definition: The expected delegation from a state with a known PVI is equal to 50%+2*PVI, so for example a D+5 state would be expected to have a delegation of 60% Democrats. Studies (e.g. Goeddert 2014) show that for every 1% shift in the national vote share there is an average shift by 2% in the number of congressional seats. Extending that to individual states, one can predict that in a 50-50 national election, a state delegation should have a Democratic fraction equal to 50% + 2*(state PVI). The percent difference between the Democratic and Republican fractions is then 4*(state PVI). The expected delegation difference between the Democratic delegation size and the Republican delegation size should be 4*(state PVI/100)*(size of the delegation), where the division by 100 is to remove the percent.

Item 4: SKEW measures the partisan fairness of a plan. Count 0 for each highly competitive district, +1 for each competitive or uncompetitive Democratic district, and -1 for each competitive or uncompetitive Republican district. Take the total for all districts in the state and subtract the expected delegation difference. Express a negative number as a positive number in favor of the Republicans. That positive number is the SKEW score, and lower numbers are closer to the ideal partisan fairness.

Item 5: POLARIZATION measures the competitiveness of a plan. Count 0 for each highly competitive district, 1 for each competitive district, and 2 for each uncompetitive district in a plan. The total for the whole state is the POLARIZATION score, and lower numbers indicate greater competitiveness.