Thoughts on a fair metric

[President Punxsutawney Phil]

Jefferson+Bibb can be a CD within quota.

Muon2 and myself would likely say that the two counties are not connected.

I would accept the path down I-20 and Alabama as a direct route, even though it nicks the corner of Tuscaloosa. Muon2 would not, unless there is a way to sneak across the Jefferson-Bibb line. The alternative route down I-65 and across on Alabama 25 is clearly through Shelby County.

But the boundary would fail my near-corner test:

Length_Gap / sqrt(Smaller_Area) > 15%.

2.2 / sqrt(658) = 8.9%

I would make an exception based on local sentiment or concentration of population. For example, about 1/3 of the Marengo County population lives in Demopolis which is at the junction of the Black Warrior river into the Tombigbee.

the two counties are contiguous by land though...

As jimrtex implies, contiguous is not the same as connected. States require contiguity, but if there's no convenient way to get from one to the other without leaving the district then they aren't connected. As a professional district drawer told me on 2011, using contiguity without internal connections is an excellent tool for gerrymandering. Therefore I prefer districts drawn to a stronger standard.

jimrtex gave you his formulation of connected. Mine is based on highways, since a road connection is a suggestion that there is an economic community of interest between places. This is from the muon rules.

You must be logged in to read this quote.

For counties the nodes are the county seats and they must be locally connected. If they are not also regionally connected there is a cost in erosity.

I personally view contiguity without internal connections as inferior, but still preferable if it serves another objective(s), something worthy enough (like reducing county splits). If an alternative exists that is just as good overall, then it ought to be considered.

[muon2]

Jefferson+Bibb can be a CD within quota.

Muon2 and myself would likely say that the two counties are not connected.

I would accept the path down I-20 and Alabama as a direct route, even though it nicks the corner of Tuscaloosa. Muon2 would not, unless there is a way to sneak across the Jefferson-Bibb line. The alternative route down I-65 and across on Alabama 25 is clearly through Shelby County.

But the boundary would fail my near-corner test:

Length_Gap / sqrt(Smaller_Area) > 15%.

2.2 / sqrt(658) = 8.9%

I would make an exception based on local sentiment or concentration of population. For example, about 1/3 of the Marengo County population lives in Demopolis which is at the junction of the Black Warrior river into the Tombigbee.

the two counties are contiguous by land though...

As jimrtex implies, contiguous is not the same as connected. States require contiguity, but if there's no convenient way to get from one to the other without leaving the district then they aren't connected. As a professional district drawer told me on 2011, using contiguity without internal connections is an excellent tool for gerrymandering. Therefore I prefer districts drawn to a stronger standard.

jimrtex gave you his formulation of connected. Mine is based on highways, since a road connection is a suggestion that there is an economic community of interest between places. This is from the muon rules.

You must be logged in to read this quote.

For counties the nodes are the county seats and they must be locally connected. If they are not also regionally connected there is a cost in erosity.

I personally view contiguity without internal connections as inferior, but still preferable if it serves another objective(s), something worthy enough (like reducing county splits). If an alternative exists that is just as good overall, then it ought to be considered.

If the goal is the promotion of fair maps then there has to be a way to quantifiably measure the maps. Look at the work that was done making a formula for political fairness in the WI case. The same work has to go into all aspects of the map.

So if non-connected contiguity that preserves integrity is allowed but not preferred, how would you quantify that preference? As an example consider how I allow locally-connected counties but prefer regionally-connected counties; I adjust for that in my erosity score.

[President Punxsutawney Phil]

Jefferson+Bibb can be a CD within quota.

Muon2 and myself would likely say that the two counties are not connected.

I would accept the path down I-20 and Alabama as a direct route, even though it nicks the corner of Tuscaloosa. Muon2 would not, unless there is a way to sneak across the Jefferson-Bibb line. The alternative route down I-65 and across on Alabama 25 is clearly through Shelby County.

But the boundary would fail my near-corner test:

Length_Gap / sqrt(Smaller_Area) > 15%.

2.2 / sqrt(658) = 8.9%

I would make an exception based on local sentiment or concentration of population. For example, about 1/3 of the Marengo County population lives in Demopolis which is at the junction of the Black Warrior river into the Tombigbee.

the two counties are contiguous by land though...

As jimrtex implies, contiguous is not the same as connected. States require contiguity, but if there's no convenient way to get from one to the other without leaving the district then they aren't connected. As a professional district drawer told me on 2011, using contiguity without internal connections is an excellent tool for gerrymandering. Therefore I prefer districts drawn to a stronger standard.

jimrtex gave you his formulation of connected. Mine is based on highways, since a road connection is a suggestion that there is an economic community of interest between places. This is from the muon rules.

You must be logged in to read this quote.

For counties the nodes are the county seats and they must be locally connected. If they are not also regionally connected there is a cost in erosity.

I personally view contiguity without internal connections as inferior, but still preferable if it serves another objective(s), something worthy enough (like reducing county splits). If an alternative exists that is just as good overall, then it ought to be considered.

If the goal is the promotion of fair maps then there has to be a way to quantifiably measure the maps. Look at the work that was done making a formula for political fairness in the WI case. The same work has to go into all aspects of the map.

So if non-connected contiguity that preserves integrity is allowed but not preferred, how would you quantify that preference? As an example consider how I allow locally-connected counties but prefer regionally-connected counties; I adjust for that in my erosity score.

Perhaps it could be done via adding an additional category, with tweaks done to encourage more whole county CDs alongside?
I wonder how this would handle counties that are literally split in two by a river, without any bridges (like Montour County, PA).

[President Punxsutawney Phil]

here's what an Alabama CD plan compromised wholely of whole county CDs and keeping everything within +/-0.5% would look like.

[muon2]

If the goal is the promotion of fair maps then there has to be a way to quantifiably measure the maps. Look at the work that was done making a formula for political fairness in the WI case. The same work has to go into all aspects of the map.

So if non-connected contiguity that preserves integrity is allowed but not preferred, how would you quantify that preference? As an example consider how I allow locally-connected counties but prefer regionally-connected counties; I adjust for that in my erosity score.

Perhaps it could be done via adding an additional category, with tweaks done to encourage more whole county CDs alongside?
I wonder how this would handle counties that are literally split in two by a river, without any bridges (like Montour County, PA).

You do a good job drawing maps, and I wish you'd been with us as we put together the rules after the 2011 cycle. I suspect you wold have added some interesting insights. So I challenge you to define the actual mechanism to allow for things like the Jefferson-Bibb CD in order to keep counties whole without opening the door to gerrymandering (as happened with the rules used in MI).

For example I think the yellow, and maybe the dark blue, CD in your map below are far more erose than desirable. The purple one isn't great either, but if the goal was to have a black performing CD then that would justify its erosity. Would you have county integrity trump all other considerations?

here's what an Alabama CD plan compromised wholely of whole county CDs and keeping everything within +/-0.5% would look like.

[President Punxsutawney Phil]

If the goal is the promotion of fair maps then there has to be a way to quantifiably measure the maps. Look at the work that was done making a formula for political fairness in the WI case. The same work has to go into all aspects of the map.

So if non-connected contiguity that preserves integrity is allowed but not preferred, how would you quantify that preference? As an example consider how I allow locally-connected counties but prefer regionally-connected counties; I adjust for that in my erosity score.

Perhaps it could be done via adding an additional category, with tweaks done to encourage more whole county CDs alongside?
I wonder how this would handle counties that are literally split in two by a river, without any bridges (like Montour County, PA).

You do a good job drawing maps, and I wish you'd been with us as we put together the rules after the 2011 cycle. I suspect you wold have added some interesting insights. So I challenge you to define the actual mechanism to allow for things like the Jefferson-Bibb CD in order to keep counties whole without opening the door to gerrymandering (as happened with the rules used in MI).

For example I think the yellow, and maybe the dark blue, CD in your map below are far more erose than desirable. The purple one isn't great either, but if the goal was to have a black performing CD then that would justify its erosity. Would you have county integrity trump all other considerations?

here's what an Alabama CD plan compromised wholely of whole county CDs and keeping everything within +/-0.5% would look like.

Thanks for the kind words.
As for your question: the most effective way to do it is probably to introduce some set of metrics that encourages both proportionality (to be defined more or less loosely depending on the area), and county integrity. In essence, limiting the leeway available to gerrymander in the first place.

[muon2]

Thanks for the kind words.
As for your question: the most effective way to do it is probably to introduce some set of metrics that encourages both proportionality (to be defined more or less loosely depending on the area), and county integrity. In essence, limiting the leeway available to gerrymander in the first place.

Right. What I'm challenging you to do is to write down those metrics. You can even start with jimrtex's or mine and modify them where you see fit. It was the exercise of writing those metrics 5-6 years ago and testing them that showed us whether they were actually workable or just a good concept.

[President Punxsutawney Phil]

Thanks for the kind words.
As for your question: the most effective way to do it is probably to introduce some set of metrics that encourages both proportionality (to be defined more or less loosely depending on the area), and county integrity. In essence, limiting the leeway available to gerrymander in the first place.

Right. What I'm challenging you to do is to write down those metrics. You can even start with jimrtex's or mine and modify them where you see fit. It was the exercise of writing those metrics 5-6 years ago and testing them that showed us whether they were actually workable or just a good concept.

My current redistricting project can be sort of a road test for this.

[muon2]

Thanks for the kind words.
As for your question: the most effective way to do it is probably to introduce some set of metrics that encourages both proportionality (to be defined more or less loosely depending on the area), and county integrity. In essence, limiting the leeway available to gerrymander in the first place.

Right. What I'm challenging you to do is to write down those metrics. You can even start with jimrtex's or mine and modify them where you see fit. It was the exercise of writing those metrics 5-6 years ago and testing them that showed us whether they were actually workable or just a good concept.

My current redistricting project can be sort of a road test for this.

But do you have written metrics first? Fitting metrics to plans after the fact is what happens in RL gerrymanders.

[President Punxsutawney Phil]

Thanks for the kind words.
As for your question: the most effective way to do it is probably to introduce some set of metrics that encourages both proportionality (to be defined more or less loosely depending on the area), and county integrity. In essence, limiting the leeway available to gerrymander in the first place.

Right. What I'm challenging you to do is to write down those metrics. You can even start with jimrtex's or mine and modify them where you see fit. It was the exercise of writing those metrics 5-6 years ago and testing them that showed us whether they were actually workable or just a good concept.

My current redistricting project can be sort of a road test for this.

But do you have written metrics first? Fitting metrics to plans after the fact is what happens in RL gerrymanders.

First, I would need to see how tightly proportionality would be defined, which effects both compactness and, to a lesser extent, county splits.

[muon2]

Thanks for the kind words.
As for your question: the most effective way to do it is probably to introduce some set of metrics that encourages both proportionality (to be defined more or less loosely depending on the area), and county integrity. In essence, limiting the leeway available to gerrymander in the first place.

Right. What I'm challenging you to do is to write down those metrics. You can even start with jimrtex's or mine and modify them where you see fit. It was the exercise of writing those metrics 5-6 years ago and testing them that showed us whether they were actually workable or just a good concept.

My current redistricting project can be sort of a road test for this.

But do you have written metrics first? Fitting metrics to plans after the fact is what happens in RL gerrymanders.

First, I would need to see how tightly proportionality would be defined, which effects both compactness and, to a lesser extent, county splits.

There are lots of definitions out there, both here on Atlas and in many states and academic papers. I'm suggesting that you define it for yourself either by adopting one you like or creating your own.

When OH had its contest in 2009 they had a clear metric for everyone to consider, but it had some unintended consequences. It was revised for the 2011 contest based on what they learned in 2009. Those objective exercises helped highlight the need to reform their system, which OH did earlier this decade. Wouldn't you like to see that happen in all states?

Some of us did and a bunch of us here started with a few coarse metrics in 2012. We went through the same sort of exercises you are doing, but at the same time looked at the results through the eyes of the metric. That metric evolved as we tried it in additional exercises, but if there wasn't a metric to start we'd just wander and argue aimlessly about which map was most fair. That's the feeling I had watching the cvparty exercise - it was purely subjective, and therefore it couldn't be judged fair in any objective way.

[President Punxsutawney Phil]

Thanks for the kind words.
As for your question: the most effective way to do it is probably to introduce some set of metrics that encourages both proportionality (to be defined more or less loosely depending on the area), and county integrity. In essence, limiting the leeway available to gerrymander in the first place.

Right. What I'm challenging you to do is to write down those metrics. You can even start with jimrtex's or mine and modify them where you see fit. It was the exercise of writing those metrics 5-6 years ago and testing them that showed us whether they were actually workable or just a good concept.

My current redistricting project can be sort of a road test for this.

But do you have written metrics first? Fitting metrics to plans after the fact is what happens in RL gerrymanders.

First, I would need to see how tightly proportionality would be defined, which effects both compactness and, to a lesser extent, county splits.

There are lots of definitions out there, both here on Atlas and in many states and academic papers. I'm suggesting that you define it for yourself either by adopting one you like or creating your own.

When OH had its contest in 2009 they had a clear metric for everyone to consider, but it had some unintended consequences. It was revised for the 2011 contest based on what they learned in 2009. Those objective exercises helped highlight the need to reform their system, which OH did earlier this decade. Wouldn't you like to see that happen in all states?

Some of us did and a bunch of us here started with a few coarse metrics in 2012. We went through the same sort of exercises you are doing, but at the same time looked at the results through the eyes of the metric. That metric evolved as we tried it in additional exercises, but if there wasn't a metric to start we'd just wander and argue aimlessly about which map was most fair. That's the feeling I had watching the cvparty exercise - it was purely subjective, and therefore it couldn't be judged fair in any objective way.

I don't think the proportionality element is fleshed out enough though.

[muon2]

Thanks for the kind words.
As for your question: the most effective way to do it is probably to introduce some set of metrics that encourages both proportionality (to be defined more or less loosely depending on the area), and county integrity. In essence, limiting the leeway available to gerrymander in the first place.

Right. What I'm challenging you to do is to write down those metrics. You can even start with jimrtex's or mine and modify them where you see fit. It was the exercise of writing those metrics 5-6 years ago and testing them that showed us whether they were actually workable or just a good concept.

My current redistricting project can be sort of a road test for this.

But do you have written metrics first? Fitting metrics to plans after the fact is what happens in RL gerrymanders.

First, I would need to see how tightly proportionality would be defined, which effects both compactness and, to a lesser extent, county splits.

There are lots of definitions out there, both here on Atlas and in many states and academic papers. I'm suggesting that you define it for yourself either by adopting one you like or creating your own.

When OH had its contest in 2009 they had a clear metric for everyone to consider, but it had some unintended consequences. It was revised for the 2011 contest based on what they learned in 2009. Those objective exercises helped highlight the need to reform their system, which OH did earlier this decade. Wouldn't you like to see that happen in all states?

Some of us did and a bunch of us here started with a few coarse metrics in 2012. We went through the same sort of exercises you are doing, but at the same time looked at the results through the eyes of the metric. That metric evolved as we tried it in additional exercises, but if there wasn't a metric to start we'd just wander and argue aimlessly about which map was most fair. That's the feeling I had watching the cvparty exercise - it was purely subjective, and therefore it couldn't be judged fair in any objective way.

I don't think the proportionality element is fleshed out enough though.

Actually we've had some fairly detailed threads, particularly a couple of years ago when we were analyzing the forumla used in the WI case and comparing to the metrics we developed. I found that there are really two distinct measures of partisan fairness. One judges how fair it is if the nation were to vote evenly for the two parties - the SKEW. The other judges how responsive the districts are to swings in the popular mood so there isn't a situation like OH where seats never change hands regardless of the direction of the wave - the POLARIZATION. Both are based on a decade of data and academic papers and are explicitly defined.

[President Punxsutawney Phil]

Thanks for the kind words.
As for your question: the most effective way to do it is probably to introduce some set of metrics that encourages both proportionality (to be defined more or less loosely depending on the area), and county integrity. In essence, limiting the leeway available to gerrymander in the first place.

Right. What I'm challenging you to do is to write down those metrics. You can even start with jimrtex's or mine and modify them where you see fit. It was the exercise of writing those metrics 5-6 years ago and testing them that showed us whether they were actually workable or just a good concept.

My current redistricting project can be sort of a road test for this.

But do you have written metrics first? Fitting metrics to plans after the fact is what happens in RL gerrymanders.

First, I would need to see how tightly proportionality would be defined, which effects both compactness and, to a lesser extent, county splits.

There are lots of definitions out there, both here on Atlas and in many states and academic papers. I'm suggesting that you define it for yourself either by adopting one you like or creating your own.

When OH had its contest in 2009 they had a clear metric for everyone to consider, but it had some unintended consequences. It was revised for the 2011 contest based on what they learned in 2009. Those objective exercises helped highlight the need to reform their system, which OH did earlier this decade. Wouldn't you like to see that happen in all states?

Some of us did and a bunch of us here started with a few coarse metrics in 2012. We went through the same sort of exercises you are doing, but at the same time looked at the results through the eyes of the metric. That metric evolved as we tried it in additional exercises, but if there wasn't a metric to start we'd just wander and argue aimlessly about which map was most fair. That's the feeling I had watching the cvparty exercise - it was purely subjective, and therefore it couldn't be judged fair in any objective way.

I don't think the proportionality element is fleshed out enough though.

Actually we've had some fairly detailed threads, particularly a couple of years ago when we were analyzing the forumla used in the WI case and comparing to the metrics we developed. I found that there are really two distinct measures of partisan fairness. One judges how fair it is if the nation were to vote evenly for the two parties - the SKEW. The other judges how responsive the districts are to swings in the popular mood so there isn't a situation like OH where seats never change hands regardless of the direction of the wave - the POLARIZATION. Both are based on a decade of data and academic papers and are explicitly defined.

The heart of my proportionality metric (as it exists in my head, a bit undefined) is probably differently defined, no? And it's probably different in results - it produced 3 Democrats in TN as opposed to 2, and 4 Democrats in Indiana as opposed to 3.
The biggest element though is looking at the presidential results and purely translating that into a number. It's what I shoot for provided the map can be clean enough. If a clean enough pure proportionality map is impossible/too unclear though, it won't be attempted.

[muon2]

Actually we've had some fairly detailed threads, particularly a couple of years ago when we were analyzing the forumla used in the WI case and comparing to the metrics we developed. I found that there are really two distinct measures of partisan fairness. One judges how fair it is if the nation were to vote evenly for the two parties - the SKEW. The other judges how responsive the districts are to swings in the popular mood so there isn't a situation like OH where seats never change hands regardless of the direction of the wave - the POLARIZATION. Both are based on a decade of data and academic papers and are explicitly defined.

The heart of my proportionality metric (as it exists in my head, a bit undefined) is probably differently defined, no? And it's probably different in results - it produced 3 Democrats in TN as opposed to 2, and 4 Democrats in Indiana as opposed to 3.
The biggest element though is looking at the presidential results and purely translating that into a number. It's what I shoot for provided the map can be clean enough. If a clean enough pure proportionality map is impossible/too unclear though, it won't be attempted.

There are multiple studies that say for every percentage point a state varies from 50-50 the delegation should vary by two points. The mistake many make is trying to get the delegation to exactly match the statewide outcome, but that's actually not fair for single member districts.

For instance IN is R+9 which means in an average election it will vote 59-41 for the Pub. With 9 seats, the expected delegation is not 59%*9 = 5.3 R to 3.7 D, but 68%*9 =  6.1 R to 2.9 D, because 18% over 50% is what is expected for the delegation. So with single member districts a 6-3 delegation on average is a fair and proportional map. Making it into a 5-4 delegation is likely a partisan gerrymander.

There are states where the minority party is too spread out to achieve this, but that's not the case in IN. It is true in MA which is R+12 so the expected delegation is (50+24) 74%*9 = 6.7 D to 2.3 R. As many have seen it's difficult to draw one R seat in MA let alone two that would be expected from the statewide numbers. So an 8-1 and even a 9-0 map in MA might not be a partisan gerrymander.

[President Punxsutawney Phil]

Actually we've had some fairly detailed threads, particularly a couple of years ago when we were analyzing the forumla used in the WI case and comparing to the metrics we developed. I found that there are really two distinct measures of partisan fairness. One judges how fair it is if the nation were to vote evenly for the two parties - the SKEW. The other judges how responsive the districts are to swings in the popular mood so there isn't a situation like OH where seats never change hands regardless of the direction of the wave - the POLARIZATION. Both are based on a decade of data and academic papers and are explicitly defined.

The heart of my proportionality metric (as it exists in my head, a bit undefined) is probably differently defined, no? And it's probably different in results - it produced 3 Democrats in TN as opposed to 2, and 4 Democrats in Indiana as opposed to 3.
The biggest element though is looking at the presidential results and purely translating that into a number. It's what I shoot for provided the map can be clean enough. If a clean enough pure proportionality map is impossible/too unclear though, it won't be attempted.

There are multiple studies that say for every percentage point a state varies from 50-50 the delegation should vary by two points. The mistake many make is trying to get the delegation to exactly match the statewide outcome, but that's actually not fair for single member districts.

For instance IN is R+9 which means in an average election it will vote 59-41 for the Pub. With 9 seats, the expected delegation is not 59%*9 = 5.3 R to 3.7 D, but 68%*9 =  6.1 R to 2.9 D, because 18% over 50% is what is expected for the delegation. So with single member districts a 6-3 delegation on average is a fair and proportional map. Making it into a 5-4 delegation is likely a partisan gerrymander.

There are states where the minority party is too spread out to achieve this, but that's not the case in IN. It is true in MA which is R+12 so the expected delegation is (50+24) 74%*9 = 6.7 D to 2.3 R. As many have seen it's difficult to draw one R seat in MA let alone two that would be expected from the statewide numbers. So an 8-1 and even a 9-0 map in MA might not be a partisan gerrymander.

My method is get the delegation to match the statewide result if that is feasible, but also balancing that against the need for compact seats and reducing county splits.
It's perfectly possible to have a delegation matching the state overall in terms of partisan balance, in many, many states. The cases where one can't? Those will balance each other out.

[muon2]

Actually we've had some fairly detailed threads, particularly a couple of years ago when we were analyzing the forumla used in the WI case and comparing to the metrics we developed. I found that there are really two distinct measures of partisan fairness. One judges how fair it is if the nation were to vote evenly for the two parties - the SKEW. The other judges how responsive the districts are to swings in the popular mood so there isn't a situation like OH where seats never change hands regardless of the direction of the wave - the POLARIZATION. Both are based on a decade of data and academic papers and are explicitly defined.

The heart of my proportionality metric (as it exists in my head, a bit undefined) is probably differently defined, no? And it's probably different in results - it produced 3 Democrats in TN as opposed to 2, and 4 Democrats in Indiana as opposed to 3.
The biggest element though is looking at the presidential results and purely translating that into a number. It's what I shoot for provided the map can be clean enough. If a clean enough pure proportionality map is impossible/too unclear though, it won't be attempted.

There are multiple studies that say for every percentage point a state varies from 50-50 the delegation should vary by two points. The mistake many make is trying to get the delegation to exactly match the statewide outcome, but that's actually not fair for single member districts.

For instance IN is R+9 which means in an average election it will vote 59-41 for the Pub. With 9 seats, the expected delegation is not 59%*9 = 5.3 R to 3.7 D, but 68%*9 =  6.1 R to 2.9 D, because 18% over 50% is what is expected for the delegation. So with single member districts a 6-3 delegation on average is a fair and proportional map. Making it into a 5-4 delegation is likely a partisan gerrymander.

There are states where the minority party is too spread out to achieve this, but that's not the case in IN. It is true in MA which is R+12 so the expected delegation is (50+24) 74%*9 = 6.7 D to 2.3 R. As many have seen it's difficult to draw one R seat in MA let alone two that would be expected from the statewide numbers. So an 8-1 and even a 9-0 map in MA might not be a partisan gerrymander.

My method is get the delegation to match the statewide result if that is feasible, but also balancing that against the need for compact seats and reducing county splits.
It's perfectly possible to have a delegation matching the state overall in terms of partisan balance, in many, many states. The cases where one can't? Those will balance each other out.

Here the math and data disagree with you. In states that that have a PVI close to even of course one should have a balanced delegation with enough competitive seats to swing with the political mood. However, once one gets to PVI 5 or more a mathematically proportional delegation is generally going to be a gerrymander. That's not a fair map. The solution is not to gerrymander single member districts but to change the law to use multi-member proportional districts.

[President Punxsutawney Phil]

Actually we've had some fairly detailed threads, particularly a couple of years ago when we were analyzing the forumla used in the WI case and comparing to the metrics we developed. I found that there are really two distinct measures of partisan fairness. One judges how fair it is if the nation were to vote evenly for the two parties - the SKEW. The other judges how responsive the districts are to swings in the popular mood so there isn't a situation like OH where seats never change hands regardless of the direction of the wave - the POLARIZATION. Both are based on a decade of data and academic papers and are explicitly defined.

The heart of my proportionality metric (as it exists in my head, a bit undefined) is probably differently defined, no? And it's probably different in results - it produced 3 Democrats in TN as opposed to 2, and 4 Democrats in Indiana as opposed to 3.
The biggest element though is looking at the presidential results and purely translating that into a number. It's what I shoot for provided the map can be clean enough. If a clean enough pure proportionality map is impossible/too unclear though, it won't be attempted.

There are multiple studies that say for every percentage point a state varies from 50-50 the delegation should vary by two points. The mistake many make is trying to get the delegation to exactly match the statewide outcome, but that's actually not fair for single member districts.

For instance IN is R+9 which means in an average election it will vote 59-41 for the Pub. With 9 seats, the expected delegation is not 59%*9 = 5.3 R to 3.7 D, but 68%*9 =  6.1 R to 2.9 D, because 18% over 50% is what is expected for the delegation. So with single member districts a 6-3 delegation on average is a fair and proportional map. Making it into a 5-4 delegation is likely a partisan gerrymander.

There are states where the minority party is too spread out to achieve this, but that's not the case in IN. It is true in MA which is R+12 so the expected delegation is (50+24) 74%*9 = 6.7 D to 2.3 R. As many have seen it's difficult to draw one R seat in MA let alone two that would be expected from the statewide numbers. So an 8-1 and even a 9-0 map in MA might not be a partisan gerrymander.

My method is get the delegation to match the statewide result if that is feasible, but also balancing that against the need for compact seats and reducing county splits.
It's perfectly possible to have a delegation matching the state overall in terms of partisan balance, in many, many states. The cases where one can't? Those will balance each other out.

Here the math and data disagree with you. In states that that have a PVI close to even of course one should have a balanced delegation with enough competitive seats to swing with the political mood. However, once one gets to PVI 5 or more a mathematically proportional delegation is generally going to be a gerrymander. That's not a fair map. The solution is not to gerrymander single member districts but to change the law to use multi-member proportional districts.

Gerrymandering to increase the likelihood of a proportionate delegation generally furthers fairness, rather than hindering it. What you are looking for is "non-partisan"; non-partisan and fair have different meanings.

[muon2]

Actually we've had some fairly detailed threads, particularly a couple of years ago when we were analyzing the forumla used in the WI case and comparing to the metrics we developed. I found that there are really two distinct measures of partisan fairness. One judges how fair it is if the nation were to vote evenly for the two parties - the SKEW. The other judges how responsive the districts are to swings in the popular mood so there isn't a situation like OH where seats never change hands regardless of the direction of the wave - the POLARIZATION. Both are based on a decade of data and academic papers and are explicitly defined.

The heart of my proportionality metric (as it exists in my head, a bit undefined) is probably differently defined, no? And it's probably different in results - it produced 3 Democrats in TN as opposed to 2, and 4 Democrats in Indiana as opposed to 3.
The biggest element though is looking at the presidential results and purely translating that into a number. It's what I shoot for provided the map can be clean enough. If a clean enough pure proportionality map is impossible/too unclear though, it won't be attempted.

There are multiple studies that say for every percentage point a state varies from 50-50 the delegation should vary by two points. The mistake many make is trying to get the delegation to exactly match the statewide outcome, but that's actually not fair for single member districts.

For instance IN is R+9 which means in an average election it will vote 59-41 for the Pub. With 9 seats, the expected delegation is not 59%*9 = 5.3 R to 3.7 D, but 68%*9 =  6.1 R to 2.9 D, because 18% over 50% is what is expected for the delegation. So with single member districts a 6-3 delegation on average is a fair and proportional map. Making it into a 5-4 delegation is likely a partisan gerrymander.

There are states where the minority party is too spread out to achieve this, but that's not the case in IN. It is true in MA which is R+12 so the expected delegation is (50+24) 74%*9 = 6.7 D to 2.3 R. As many have seen it's difficult to draw one R seat in MA let alone two that would be expected from the statewide numbers. So an 8-1 and even a 9-0 map in MA might not be a partisan gerrymander.

My method is get the delegation to match the statewide result if that is feasible, but also balancing that against the need for compact seats and reducing county splits.
It's perfectly possible to have a delegation matching the state overall in terms of partisan balance, in many, many states. The cases where one can't? Those will balance each other out.

Here the math and data disagree with you. In states that that have a PVI close to even of course one should have a balanced delegation with enough competitive seats to swing with the political mood. However, once one gets to PVI 5 or more a mathematically proportional delegation is generally going to be a gerrymander. That's not a fair map. The solution is not to gerrymander single member districts but to change the law to use multi-member proportional districts.

Gerrymandering to increase the likelihood of a proportionate delegation generally furthers fairness, rather than hindering it. What you are looking for is "non-partisan"; non-partisan and fair have different meanings.

That depends the context of the question. Gerrymandering is drawing districts to get a particular partisan outcome. I and many others would say that gerrymandering is inherently unfair, even if the partisan goals are laudable. The problem is that shifts in the electorate can change enough within a decade that the partisan goals are no longer met, and then the unfairness become more obvious. If you look at some old threads I suspect you can find plans drawn early in the decade that were thought to meet proportionality but by 2018 would look like nothing but a mild partisan gerrymander.

[President Punxsutawney Phil]

Actually we've had some fairly detailed threads, particularly a couple of years ago when we were analyzing the forumla used in the WI case and comparing to the metrics we developed. I found that there are really two distinct measures of partisan fairness. One judges how fair it is if the nation were to vote evenly for the two parties - the SKEW. The other judges how responsive the districts are to swings in the popular mood so there isn't a situation like OH where seats never change hands regardless of the direction of the wave - the POLARIZATION. Both are based on a decade of data and academic papers and are explicitly defined.

The heart of my proportionality metric (as it exists in my head, a bit undefined) is probably differently defined, no? And it's probably different in results - it produced 3 Democrats in TN as opposed to 2, and 4 Democrats in Indiana as opposed to 3.
The biggest element though is looking at the presidential results and purely translating that into a number. It's what I shoot for provided the map can be clean enough. If a clean enough pure proportionality map is impossible/too unclear though, it won't be attempted.

There are multiple studies that say for every percentage point a state varies from 50-50 the delegation should vary by two points. The mistake many make is trying to get the delegation to exactly match the statewide outcome, but that's actually not fair for single member districts.

For instance IN is R+9 which means in an average election it will vote 59-41 for the Pub. With 9 seats, the expected delegation is not 59%*9 = 5.3 R to 3.7 D, but 68%*9 =  6.1 R to 2.9 D, because 18% over 50% is what is expected for the delegation. So with single member districts a 6-3 delegation on average is a fair and proportional map. Making it into a 5-4 delegation is likely a partisan gerrymander.

There are states where the minority party is too spread out to achieve this, but that's not the case in IN. It is true in MA which is R+12 so the expected delegation is (50+24) 74%*9 = 6.7 D to 2.3 R. As many have seen it's difficult to draw one R seat in MA let alone two that would be expected from the statewide numbers. So an 8-1 and even a 9-0 map in MA might not be a partisan gerrymander.

My method is get the delegation to match the statewide result if that is feasible, but also balancing that against the need for compact seats and reducing county splits.
It's perfectly possible to have a delegation matching the state overall in terms of partisan balance, in many, many states. The cases where one can't? Those will balance each other out.

Here the math and data disagree with you. In states that that have a PVI close to even of course one should have a balanced delegation with enough competitive seats to swing with the political mood. However, once one gets to PVI 5 or more a mathematically proportional delegation is generally going to be a gerrymander. That's not a fair map. The solution is not to gerrymander single member districts but to change the law to use multi-member proportional districts.

Gerrymandering to increase the likelihood of a proportionate delegation generally furthers fairness, rather than hindering it. What you are looking for is "non-partisan"; non-partisan and fair have different meanings.

That depends the context of the question. Gerrymandering is drawing districts to get a particular partisan outcome. I and many others would say that gerrymandering is inherently unfair, even if the partisan goals are laudable. The problem is that shifts in the electorate can change enough within a decade that the partisan goals are no longer met, and then the unfairness become more obvious. If you look at some old threads I suspect you can find plans drawn early in the decade that were thought to meet proportionality but by 2018 would look like nothing but a mild partisan gerrymander.

I do not share the opinion that gerrymandering is inherently unfair.
As for shifts in voting and in the electorate? That might be solved by loosening the threshold depending on the elasticity of the area in question...

[muon2]

I split this so Monarch can have the original thread back.