6 ms·

Utility Ghost: Gamified redistricting with partisan symmetry

80 points by mathfan 8y ago

nscalf 8y ago
So effectively, both parties take turns choosing what "atoms" (counties, districts, etc) belong in which districts. They must be contiguous. I'd like to see how this relates to the efficiency gap that was heavily used around gerrymandering in Wisconsin. It's mentioned, but they don't go back to analyze how it would be---at least I didn't see it from a simple search.
- qfwfq_ 8y ago
  The efficiency gap is just one measure of partisan bias that can be used to assess the utility/efficacy of any redistricting plan. The game the authors propose is (basically) an algorithm to draw good districts between two players. So, there's no real or expected relationship between the two... partisan control (the objective the authors cite) isn't necessarily related to the efficiency of that control measured by the efficiency gap. The problem is that so many different measures of fairness exist, have validity, and can be used in different legal or jurisprudential contexts, we're always talking past each other on the meaning of these measures.
brianolson 8y ago
The abstract describes a system for achieving shoddy Proportional Representation through bipartisan (or multi-partisan) gerrymandering. If you want Proportional Representation, then just do that and don't try to gerrymander districts to achieve it. I think this country is too stuck on the notion of districts and needs to think outside the box and paint outside the lines (nyuks intended). I think we keep describing what we want as some sort of identity/ideology based representation and those things are now much more important to us than geography and _where_ a person lives. So we should have at-large PR or large multi-member districts (8 or more reps per district).
- TheLoneAdmin 8y ago
  I agree we should switch to Proportional Representation. Unfortunately this would weak our duopoly parties, allowing third parties to win seats. So they (D&R) will never change the laws. In a FPTP election, 10-20% of the votes are wasted on a 3rd party. In a PR election, 10-20% of the votes get 10-20% of the seats.
  - dragonwriter 8y ago
    > Unfortunately this would weak our duopoly parties, allowing third parties to win seats. So they (D&R) will never change the laws. (1) They will change the law if the choice is being displaced as major parties (i.e., if people care enough), and (2) for state government, in many cases the elected members of political parties would not need to act, the citizenry could change the laws directly, > In a FPTP election, 10-20% of the votes are wasted on a 3rd party. In a PR election, 10-20% of the votes get 10-20% of the seats. Assuming you mean party list PR, that's true, but then also individual candidates aren't accountable to the general electorate and no one has anyone they can point to as “their” representative. That's a significant trade-off; there are systems which achieve rough proportionality (far better than FPTP) that don't have these problems (i.e., STV with small, e.g. 5-member, districts), but if you look simply at a naive “wastee vote” metric as you've constructed here they may not be better than FPTP. > In a FPTP election, 10-20% of the votes are wasted on a 3rd party In the US, the number is much lower; of course, the problem is as big or bigger if you add in abstentions and tactical votes.
    - brianolson 8y ago
      I advocate STV with larger sets, 8 or more. The wasted vote is smaller, any constituency that can show up and get out the vote for 1/N of the total gets one of the N seats. 5 seats or fewer biases towards the major parties. Lots of 5 seat sets will go 3-to-2 or 2-to-3. I'd like to get in other voices and split it 5/3/1 or 4/4/1. And I bet over time that 1 would grow to 2, but we have a bootstrapping problem in the current system that always throws away that smaller voice.
      - dragonwriter 8y ago
- mathfan 8y ago
  A switch to PR would require a change to federal law. Meanwhile, in the absence of PR, any state is free to adopt its own redistricting policy. The latter approach seems much more feasible.
  - brianolson 8y ago
    A State changing how it runs its State Legislature is a change to the State Constitution either way, whether taking away the map drawing authority from the Legislature itself or changing the structure of the Legislature to be non-districted or multi-member districted.
    - mathfan 8y ago
      What is more likely, a change to federal law, or the existence of one of 50 states that changes state law?
  - ende 8y ago
    It’s a single statute that requires single member districts. It wouldn’t be difficult to change. Hell, it may even be unconstitutional.
    - dragonwriter 8y ago
      > It’s a single statute that requires single member districts. It wouldn’t be difficult to change. Yes, it would, politically. The reason it exists will produce vet strong resistance to a straight repeal, and any replacement with an alternate approach to preventing the original problem while allowing multimember districts will, in addition to resistance from those in Congress whose opposition is driven by opposing undermining the existing dominant parties, resistance around the details of any alternative scheme (which will have predictable winners and losers.)
- Arnt 8y ago
  I wonder if gerrymandering even makes sense with the current Swiss way of counting? The way they count is unorthodox: It needs a computer to compute the result because it's very iterative, but the result is easily verified without a computer. In Condorcet's days this would have been unthinkable. Each party receives the right number of seats based on number of votes, and the iterations progressively move seats around to favour candidates who are popular in their party and in their region/district. Eventually it says "can't be made more fair now" and at that point it's easy to verify that. At first blush it seems that gerrymandering could then only have an effect within a party... is that right?
- ihaveajob 8y ago
  Coming from a place with proportional distribution, I can see there's value in a winnner-takes-all policy. For example, alternance is more likely (in Andalusia it took 37 years for a non-socialist government to have a chance at power). Also, representatives respond to their district's specific needs, which is good because you have one specific person to go talk to, and to boot off their seats if necessary (unlike a system where the politicians at the top of their parties ranked lists are pretty much guaranteed a position). There's no perfect system, and the grass is always greener, etc.
dane-pgp 8y ago
It occurred to me that rather than trying to fix the gerrymandering problem at the map-drawing stage, it might be simpler to solve it at the election stage, where the imbalance is more obvious. My proposal, which I call "merrymandering", involves comparing the number of seats won by each party in a state after an election, with the aggregate percentage of votes won by each party in that election. If there is an imbalance, then the over-represented party has one of their seats assigned to the next-most-over-represented party, and so on until any seat change would result in a more disproportionate result. The choice of which seat gets reassigned could be chosen based on how close the other party came to winning that seat, to make it deterministic, or a random process could be used, to avoid safe seats. In practice, what I think would happen is that with this system in place, there would be no advantage to partisan redistricting, so the merrymandering ruleset would never actually be applied, and no seats would flip. Nevertheless, I have some confidence that the system would survive a legal challenge, since it only changes election results to make them strictly more compliant with the "one person one vote" principle.
- iainmerrick 8y ago
  I’ve had a similar idea: assign a weight to each party, and multiply each vote by the appropriate weight. In standard FPTP, all the weights are one. To make the results proportional, just assign weights such that the seat counts match the overall vote shares (rounding in favor of the parties with the most votes). Edit to add: I would guess there’s probably a single solution for any given election, or rather a contiguous set of solutions, but I haven’t verified that. It probably wouldn’t survive a legal challenge, as some would complain that votes are being counted unequally. Though in this situation I’d argue that the ends (PR) justify the means (unequal weights).
  - dane-pgp 8y ago
    I like this idea as well, and first encountered it at this site: http://www.dprvoting.org/ http://www.dprvoting.org/ Their system is called "Direct Party and Representative Voting" (DPR Voting) and it has some weird edge cases (like dealing with independent candidates, or parties that win a big vote share but no actual seats), but it definitely seems like an improvement over FPTP, and uses a very easy/quick ballot counting process (unlike IRV or Approval voting, for example).
madrox 8y ago
I'm not convinced how we elect representatives is a problem. Gerrymandering seems an annoying an unfair way to elevate certain people, sure, but I don't think fixing this is the panacea to toxic politics that everyone thinks it is. You simply have to look internationally at other systems to see they also have problems...different problems from ours, sure, but not exactly better.
- mathfan 8y ago
  The Supreme Court has already stated that partisan gerrymandering is unconstitutional, we just don't have a mechanism to systematically detect it after the fact. That's why it's so hard to fight in the courts. If possible, I think it's a good idea to prevent unconstitutional gerrymandering, even if it doesn't solve all of our problems.

80 points by mathfan 8y ago

So effectively, both parties take turns choosing what "atoms" (counties, districts, etc) belong in which districts. They must be contiguous. I'd like to see how this relates to the efficiency gap that was heavily used around gerrymandering in Wisconsin. It's mentioned, but they don't go back to analyze how it would be---at least I didn't see it from a simple search.

qfwfq_ 8y ago
The efficiency gap is just one measure of partisan bias that can be used to assess the utility/efficacy of any redistricting plan. The game the authors propose is (basically) an algorithm to draw good districts between two players. So, there's no real or expected relationship between the two... partisan control (the objective the authors cite) isn't necessarily related to the efficiency of that control measured by the efficiency gap. The problem is that so many different measures of fairness exist, have validity, and can be used in different legal or jurisprudential contexts, we're always talking past each other on the meaning of these measures.

brianolson 8y ago

The abstract describes a system for achieving shoddy Proportional Representation through bipartisan (or multi-partisan) gerrymandering. If you want Proportional Representation, then just do that and don't try to gerrymander districts to achieve it. I think this country is too stuck on the notion of districts and needs to think outside the box and paint outside the lines (nyuks intended). I think we keep describing what we want as some sort of identity/ideology based representation and those things are now much more important to us than geography and _where_ a person lives. So we should have at-large PR or large multi-member districts (8 or more reps per district).

TheLoneAdmin 8y ago
I agree we should switch to Proportional Representation. Unfortunately this would weak our duopoly parties, allowing third parties to win seats. So they (D&R) will never change the laws. In a FPTP election, 10-20% of the votes are wasted on a 3rd party. In a PR election, 10-20% of the votes get 10-20% of the seats.
- dragonwriter 8y ago
  > Unfortunately this would weak our duopoly parties, allowing third parties to win seats. So they (D&R) will never change the laws. (1) They will change the law if the choice is being displaced as major parties (i.e., if people care enough), and (2) for state government, in many cases the elected members of political parties would not need to act, the citizenry could change the laws directly, > In a FPTP election, 10-20% of the votes are wasted on a 3rd party. In a PR election, 10-20% of the votes get 10-20% of the seats. Assuming you mean party list PR, that's true, but then also individual candidates aren't accountable to the general electorate and no one has anyone they can point to as “their” representative. That's a significant trade-off; there are systems which achieve rough proportionality (far better than FPTP) that don't have these problems (i.e., STV with small, e.g. 5-member, districts), but if you look simply at a naive “wastee vote” metric as you've constructed here they may not be better than FPTP. > In a FPTP election, 10-20% of the votes are wasted on a 3rd party In the US, the number is much lower; of course, the problem is as big or bigger if you add in abstentions and tactical votes.
  - brianolson 8y ago
    I advocate STV with larger sets, 8 or more. The wasted vote is smaller, any constituency that can show up and get out the vote for 1/N of the total gets one of the N seats. 5 seats or fewer biases towards the major parties. Lots of 5 seat sets will go 3-to-2 or 2-to-3. I'd like to get in other voices and split it 5/3/1 or 4/4/1. And I bet over time that 1 would grow to 2, but we have a bootstrapping problem in the current system that always throws away that smaller voice.
    - dragonwriter 8y ago
mathfan 8y ago
A switch to PR would require a change to federal law. Meanwhile, in the absence of PR, any state is free to adopt its own redistricting policy. The latter approach seems much more feasible.
- brianolson 8y ago
  A State changing how it runs its State Legislature is a change to the State Constitution either way, whether taking away the map drawing authority from the Legislature itself or changing the structure of the Legislature to be non-districted or multi-member districted.
  - mathfan 8y ago
    What is more likely, a change to federal law, or the existence of one of 50 states that changes state law?
- ende 8y ago
  It’s a single statute that requires single member districts. It wouldn’t be difficult to change. Hell, it may even be unconstitutional.
  - dragonwriter 8y ago
    > It’s a single statute that requires single member districts. It wouldn’t be difficult to change. Yes, it would, politically. The reason it exists will produce vet strong resistance to a straight repeal, and any replacement with an alternate approach to preventing the original problem while allowing multimember districts will, in addition to resistance from those in Congress whose opposition is driven by opposing undermining the existing dominant parties, resistance around the details of any alternative scheme (which will have predictable winners and losers.)
Arnt 8y ago
I wonder if gerrymandering even makes sense with the current Swiss way of counting? The way they count is unorthodox: It needs a computer to compute the result because it's very iterative, but the result is easily verified without a computer. In Condorcet's days this would have been unthinkable. Each party receives the right number of seats based on number of votes, and the iterations progressively move seats around to favour candidates who are popular in their party and in their region/district. Eventually it says "can't be made more fair now" and at that point it's easy to verify that. At first blush it seems that gerrymandering could then only have an effect within a party... is that right?
ihaveajob 8y ago
Coming from a place with proportional distribution, I can see there's value in a winnner-takes-all policy. For example, alternance is more likely (in Andalusia it took 37 years for a non-socialist government to have a chance at power). Also, representatives respond to their district's specific needs, which is good because you have one specific person to go talk to, and to boot off their seats if necessary (unlike a system where the politicians at the top of their parties ranked lists are pretty much guaranteed a position). There's no perfect system, and the grass is always greener, etc.

dane-pgp 8y ago

It occurred to me that rather than trying to fix the gerrymandering problem at the map-drawing stage, it might be simpler to solve it at the election stage, where the imbalance is more obvious. My proposal, which I call "merrymandering", involves comparing the number of seats won by each party in a state after an election, with the aggregate percentage of votes won by each party in that election. If there is an imbalance, then the over-represented party has one of their seats assigned to the next-most-over-represented party, and so on until any seat change would result in a more disproportionate result. The choice of which seat gets reassigned could be chosen based on how close the other party came to winning that seat, to make it deterministic, or a random process could be used, to avoid safe seats. In practice, what I think would happen is that with this system in place, there would be no advantage to partisan redistricting, so the merrymandering ruleset would never actually be applied, and no seats would flip. Nevertheless, I have some confidence that the system would survive a legal challenge, since it only changes election results to make them strictly more compliant with the "one person one vote" principle.

iainmerrick 8y ago
I’ve had a similar idea: assign a weight to each party, and multiply each vote by the appropriate weight. In standard FPTP, all the weights are one. To make the results proportional, just assign weights such that the seat counts match the overall vote shares (rounding in favor of the parties with the most votes). Edit to add: I would guess there’s probably a single solution for any given election, or rather a contiguous set of solutions, but I haven’t verified that. It probably wouldn’t survive a legal challenge, as some would complain that votes are being counted unequally. Though in this situation I’d argue that the ends (PR) justify the means (unequal weights).
- dane-pgp 8y ago
  I like this idea as well, and first encountered it at this site: http://www.dprvoting.org/ http://www.dprvoting.org/ Their system is called "Direct Party and Representative Voting" (DPR Voting) and it has some weird edge cases (like dealing with independent candidates, or parties that win a big vote share but no actual seats), but it definitely seems like an improvement over FPTP, and uses a very easy/quick ballot counting process (unlike IRV or Approval voting, for example).

madrox 8y ago

I'm not convinced how we elect representatives is a problem. Gerrymandering seems an annoying an unfair way to elevate certain people, sure, but I don't think fixing this is the panacea to toxic politics that everyone thinks it is. You simply have to look internationally at other systems to see they also have problems...different problems from ours, sure, but not exactly better.

mathfan 8y ago
The Supreme Court has already stated that partisan gerrymandering is unconstitutional, we just don't have a mechanism to systematically detect it after the fact. That's why it's so hard to fight in the courts. If possible, I think it's a good idea to prevent unconstitutional gerrymandering, even if it doesn't solve all of our problems.