Based on 2012 results only, which state gerrymander flipped the most seats?

[Torie]

I am not persuaded that the rules I propose would systematically reduce the number of competitive districts, or increase the skew for that matter. That is your claim, but so far, I don't see any evidence of that. What my rules I think would do is reduce chop penalty scores. There is no flexibility in the sense that it involves humans. It involves a computer. A computer does not have a partisan bias, unless programmed for it.

I do have a preference for UCC integrity, but if after the penalty points, another map is superior, as your Colorado map was, splendid. It takes a bit more creativity to do what you did, because if you hew to UCC integrity, the remaining options are fewer.

You avoided both my questions. This is not merely a computer exercise, but one that could be applied as a crowdsourcing activity. It's as a way to engage the public in a constrained fashion that I envisage a set of rules. I'll restate my questions.

Question one. You used to hold that if all else is equal, competitive districts were better than uncompetitive districts. You called plans with more competitive districts "the cat's meow" IIRC. Do you still feel that way?

Yes.

Question two. On one hand you prefer UCC as an objective proxy for a CoI, and the UCC was not a byproduct of other rules. On the other hand in the context of connections you have stated that you are not concerned with CoI except as a byproduct of other objective rules. Should there be specific objective rules designed solely as proxies for CoI?

The UCC matter is about rural versus urban. I think that is a popular and agreed upon COI metric, unlike most of them, which are BS, and used for partisan reasons. But obviously, chops and erosity have something to do with COI. Boards of Elections don't like chops, and erose maps will be the subject of public derision.  In addition, without the UCC rules, there probably would be maps with very different partisan complexions, but equal scores, and I don't think that is a good thing. I don't think this highway thing, and whether a local highway, or a state highway, is very useful as a COI proxy, and I don't think anybody really cares.

Oh, I don't know what "crowd sourcing" means here. It is a computer exercise. Granted maybe the programming of the computer might not be perfect,, and somebody might be able to beat it. But no human judgment is involved, and thus there is no game playing involved.

[jimrtex]

Here's my take on CO. I reduced some erosity by swapping a pack point for a county chop. I also avoided a questionable link between Gilpin and Clear Lake. The districts are 2D, 1d, 1e, 3R compared to the actual 2D, 1d, 1e, 1r, 2R. Yes the 1e is R+0 in mine compared to D+1, but I don't see a substantial change.

That is a four lane highway built by Central City to provide access from Clear Creek County (I-70 and Denver really).

And that's an area where our definitions conflict. The Central City Parkway is neither numbered nor a state highway, though it was built locally to provide a more convenient and direct connection. I can imagine at some future point the state taking over the highway, at which point it would meet both our definitions. I'm willing to miss a few like this in the interest of a simpler definition.

In any case I went for lower erosity for the same chop count.

How would someone drive throughout your proposed district?

[jimrtex]

You avoided both my questions. This is not merely a computer exercise, but one that could be applied as a crowdsourcing activity. It's as a way to engage the public in a constrained fashion that I envisage a set of rules. I'll restate my questions.

Oh, I don't know what "crowd sourcing" means here. It is a computer exercise. Granted maybe the programming of the computer might not be perfect,, and somebody might be able to beat it. But no human judgment is involved, and thus there is no game playing involved.

Why do you think it is a computer exercise?

I envision the public submitting plans that conform to certain constraints. These would be graded, likely by a computer.

The best plans would be voted on by (a sample of) the public.

[Torie]

You avoided both my questions. This is not merely a computer exercise, but one that could be applied as a crowdsourcing activity. It's as a way to engage the public in a constrained fashion that I envisage a set of rules. I'll restate my questions.

Oh, I don't know what "crowd sourcing" means here. It is a computer exercise. Granted maybe the programming of the computer might not be perfect,, and somebody might be able to beat it. But no human judgment is involved, and thus there is no game playing involved.

Why do you think it is a computer exercise?

I envision the public submitting plans that conform to certain constraints. These would be graded, likely by a computer.

The best plans would be voted on by (a sample of) the public.

I had not heard of this "vote" by the public before. Anywhere, the system by which a plan is selected after the maps are scored by a computer (and I don't see why a computer cannot just generate the maps itself, since there is no human discretion involved, albeit maybe the programming will not be perfect and the map can be beaten), is another issue. I have my own scheme for that that I have described.

[jimrtex]

You avoided both my questions. This is not merely a computer exercise, but one that could be applied as a crowdsourcing activity. It's as a way to engage the public in a constrained fashion that I envisage a set of rules. I'll restate my questions.

Oh, I don't know what "crowd sourcing" means here. It is a computer exercise. Granted maybe the programming of the computer might not be perfect,, and somebody might be able to beat it. But no human judgment is involved, and thus there is no game playing involved.

Why do you think it is a computer exercise?

I envision the public submitting plans that conform to certain constraints. These would be graded, likely by a computer.

The best plans would be voted on by (a sample of) the public.

I had not heard of this "vote" by the public before. Anywhere, the system by which a plan is selected after the maps are scored by a computer (and I don't see why a computer cannot just generate the maps itself, since there is no human discretion involved, albeit maybe the programming will not be perfect and the map can be beaten), is another issue. I have my own scheme for that that I have described.

AFAIK, Muon2 and I have always seen this as redistricting by crowd-sourcing.  We were treating you as a model for a publicly-submitted plan. We have not been describing the parameters of a computer program, but rather constraints on publicly submitted maps.

In most instances where there is public input of whole maps it is ignored. It is either in conflict with the ideas of the map drawers, be they the legislature or a commission, or it is too complicated to mesh them together.

A particular reason for emphasizing whole counties is that it is not only good redistricting practice, but it is more amenable to public input.

If we simply let people submit maps, the legislature or commission will thank the public for their effort and public interest in this important issue, then throw the maps into a dumpster.

But if there is some way to measure "better" plans - and I believe that there are - someone must pick the best plan. Small groups are subject to bias.

But we let the public choose the representative for their district, why shouldn't we let them choose their district?

[muon2]

When I originated my rules it was in response to the OH redistricting competition (2011). That was a crowdsourced exercise in developing a map. Rules were developed and posted, then tools were provided to the public to make submissions. As plans were submitted others could take those plans and create modified versions of the same. At the end of the public exercise the winning plans were selected and presented to the legislature. Both jimrtex and I participated in that competition and we had multiple threads about it at the time. Afterwards, there were aspects of the OH competition rules that I thought could be improved and generalized.

The current battle in FL uses some of that crowdsourcing idea. There is a public mapping tool and anyone can get an account and submit plans. A plan can be completely new or based on plans already submitted. They score certain aspects based on the FL constitution. It's all public, and jimrtex even contributed a plan to the process.

The OH competition included a plan generated by a computer algorithm, but it did not do as well as the best plans submitted by the public. The 2011 OH competition forced a single winner with specific second and third places, but that required a set of weights on the disparate measures for geographic integrity, compactness, and partisan fairness. Those weights were arbitrary and forced more attention on some measures than others. The 2009 test run for the competition in OH kept the factors separate and looked at the best in each category. I turned that idea into a Pareto test based on separate measures.

What I would prefer is a crowdsourced map exercise with predetermined rules using a publicly available mapping tool with data. Anyone could participate including computer algorithms and partisan interests. At the end of the exercise the set of plans on the Pareto frontier would go to a decision making body - either a commission or the legislature. That body would then have to select from that limited set of plans. This type of approach got the attention of the Council of State Governments which featured it in its monthly newsletter a few years ago.

I should add that the VA redistricting exercise we started on this board a year ago was built on this model and outlined the process in some detail.

[muon2]

Question two. On one hand you prefer UCC as an objective proxy for a CoI, and the UCC was not a byproduct of other rules. On the other hand in the context of connections you have stated that you are not concerned with CoI except as a byproduct of other objective rules. Should there be specific objective rules designed solely as proxies for CoI?

The UCC matter is about rural versus urban. I think that is a popular and agreed upon COI metric, unlike most of them, which are BS, and used for partisan reasons. But obviously, chops and erosity have something to do with COI. Boards of Elections don't like chops, and erose maps will be the subject of public derision.  In addition, without the UCC rules, there probably would be maps with very different partisan complexions, but equal scores, and I don't think that is a good thing. I don't think this highway thing, and whether a local highway, or a state highway, is very useful as a COI proxy, and I don't think anybody really cares.

We may think that UCC rules make sense, but I had a chance to test that on another group of interested users. In the Forum Redistricting Commission for VA that we set up a year ago, there was little interest in UCCs as an unneeded complication. They preferred just to consider chops. Based on that result, I'm not so sure I would judge them as popular and agreed upon outside of the group of us that developed them two years ago.

[Torie]

Interesting about UCC's. Anyway, there is a policy reason for them as I described. All one can do is advocate. But on the chops and erosity thing, I assume that you agree that it is also popular.

Your methodology might well be useful as a PR thing (and again, just to increase the odds that something was not missed), but in the end, again, the scores are the scores, so it is matter of what the scoring system produces as an empirical matter, and not the potential for playing games by partisan hacks. After the maps that make the cut are chosen, I like my method for determining the final map, which does not seem in conflict with your PR scheme.

[muon2]

Interesting about UCC's. Anyway, there is a policy reason as I described.

I agree that there is a policy reason for respecting UCCs. I've just never been convinced of the policy reason for the pack aspect. The cover rule is an equivalent of the chop rule, it just looks at the whole UCC rather than one county. The pack rule is equivalent to a rule that would penalize plans that don't put as many whole districts as possible into a large county. But we didn't see county packs as a necessary for good policy, and I feel the same way about UCC packs. I was outvoted by train and you so we have them.

[Torie]

Interesting about UCC's. Anyway, there is a policy reason as I described.

I agree that there is a policy reason for respecting UCCs. I've just never been convinced of the policy reason for the pack aspect. The cover rule is an equivalent of the chop rule, it just looks at the whole UCC rather than one county. The pack rule is equivalent to a rule that would penalize plans that don't put as many whole districts as possible into a large county. But we didn't see county packs as a necessary for good policy, and I feel the same way about UCC packs. I was outvoted by train and you so we have them.

Or, don't be intimidated by the vote thing. And over time, attitudes change. Mine have some at the margins. The more you do this, the more insight that one has. The policy reason for the pack rule, is that otherwise about a half CD more of urban voters could be combined with rural voters absent the pack rule, as compared with the pack rule being in place. And with that population increase in combination CD's, the odds for having maps with wider partisan ranges, increases.

[muon2]

Interesting about UCC's. Anyway, there is a policy reason as I described.

I agree that there is a policy reason for respecting UCCs. I've just never been convinced of the policy reason for the pack aspect. The cover rule is an equivalent of the chop rule, it just looks at the whole UCC rather than one county. The pack rule is equivalent to a rule that would penalize plans that don't put as many whole districts as possible into a large county. But we didn't see county packs as a necessary for good policy, and I feel the same way about UCC packs. I was outvoted by train and you so we have them.

Or, don't be intimidated by the vote thing. And over time, attitudes change. Mine have some at the margins. The more you do this, the more insight that one has. The policy reason for the pack rule, is that otherwise about a half CD more of urban voters could be combined with rural voters absent the pack rule, as compared with the pack rule being in place. And with that population increase in combination CD's, the odds for having maps with wider partisan ranges, increases.

But I would argue the same thing applies for one county dominating another, much as with your example of urban and rural. It seems inconsistent to me to be concerned about urban/rural but not large/small county. The lack of a pack rule can cut either way, and as you would note it reduces flexibility - in this case to keep whole counties.

We have only a cover rule for counties, called chops. Initially we had only a cover rule for UCCs, and that made sense to me from a consistency point of view. Then much later we added the pack rule, but never applied it back to smaller jurisdictions.

Don't get me wrong. We spent a lot of time three years ago defining the chop. And after a lot of review we decided that a pack rule for counties wasn't needed - it would as often work for a large county as it would against it. I don't think we need to revisit that.

[jimrtex]

Interesting about UCC's. Anyway, there is a policy reason as I described.

I agree that there is a policy reason for respecting UCCs. I've just never been convinced of the policy reason for the pack aspect. The cover rule is an equivalent of the chop rule, it just looks at the whole UCC rather than one county. The pack rule is equivalent to a rule that would penalize plans that don't put as many whole districts as possible into a large county. But we didn't see county packs as a necessary for good policy, and I feel the same way about UCC packs. I was outvoted by train and you so we have them.

For legislative districts where there are likely to be more counties with multiple districts, I think a pack rule usually makes sense.

Consider a county entitled to N.6 districts.   If you have N-1 whole districts, and split the remaining 1.6, 0.8 and 0.8, the large county dominates N+1 districts.  If you have N whole districts, then the remaining 0.6 will likely result in a shared district.

There might be some instances where the 0.6 would have be stretched to cross the larger county to reach two smaller of 0.25 and 0.15 but this is rare.

I was able to pack 11 of 12 largest counties in Florida.  The 12th was Miami-Dade, where the extra district was caused by Monroe, which is a rather special circumstance. Only one such split was particularly ugly, that of Brevard where I had to split Palm Bay. But if I had split Brevard between two districts that crossed into other counties, I would have ended up splitting another county as well.

[Torie]

Here's Indiana. The roll of the metric dice proved quite kind to the Democrats here. All of Portage ended up in IN-02, pushing the CD into the tossup range. The Pub PVI in IN-08 also dropped about 1.5 points, but it is still a Pub CD. So the metric's application caused the Dems to pick up half a seat here. Granted, in 2012, the Pub trend was so strong, that IN-02 moved into the Pub zone. But we are only looking at the 2008 numbers.

[Torie]

And here is Oregon. For some reason Dave's utility has no partisan data for Oregon, but I strongly suspect that OR-05 moved from tossup to Pub, so I am going to mark that the map enacted cost the Pubs half a seat. In this case, remarkably enough, the population variance between CD's ended up being less than 1,000 people. That's a first for me.

Oh, there is a bridge chop here (Wasco and Marion have no road connections whatsoever, so the state highway route is through chopped Clackamas). This is an example as to why bridge chops should be allowed, if disfavored by losing in a tie situation. I very much doubt there is another map with the same chop count.

I must say, the more I do this, the more that I admire Muon2's metrics (they are still not "perfect," but give me time. ) He should be congratulated, and very proud of himself. Map after map looks very good to me, with most far better than what was actually drawn, even if what was actually drawn was not intended to be a gerrymander. When he writes his seminal paper, and causes a press sensation, hopefully my name will appear in the acknowledgment section, along with Jimrtex of course, and Train, and all those who constructively engaged in this grand, and most important, endeavor.





Well actually, below is the winning map, because while it increases population inequality, it improves erosity by a point, which is a higher tier consideration.



Oh, my bad. The map below has the same chop count, since it "wins" the chop contest. I wonder which of the two maps above has the lower erosity score?    God bless bridge chops! Oh, there is no road link between nodes in Linn and Jefferson either. Well, just assume there were one. It's the concept that counts!

[Nyvin]

Oregon has crazy state laws regarding redistricting.     I'm pretty sure all the counties east of the Cascades are considered a "community of interest" and can't be divided by state law ORS 188.010.  

Also Portland has 2 different communities of interest and has to be split east and west....I think going along the river.

And it's in the state law that the districts have to be connected with transportation links.

[Torie]

Oregon has crazy state laws regarding redistricting.     I'm pretty sure all the counties east of the Cascades are considered a "community of interest" and can't be divided by state law ORS 188.010.  

Also Portland has 2 different communities of interest and has to be split east and west....I think going along the river.

And it's in the state law that the districts have to be connected with transportation links.

The idea is to have a uniform model code, for each state to consider to enact. That is the purpose of this exercise. And thus, each and every state, which deemed the model code a good idea, would dump its existing law into the dust bin.

[muon2]

Here's Indiana. The roll of the metric dice proved quite kind to the Democrats here. All of Portage ended up in IN-02, pushing the CD into the tossup range. The Pub PVI in IN-08 also dropped about 1.5 points, but it is still a Pub CD. So the metric's application caused the Dems to pick up half a seat here. Granted, in 2012, the Pub trend was so strong, that IN-02 moved into the Pub zone. But we are only looking at the 2008 numbers.

It looks like there are 3 county chops more than the minimum and some number of township chops in Marion. Here's an alternative that only chops Marion, and has no township chops in Marion. This application of the metric gives a 2D, 3r, 4R split with the closest district CD-2 at r+2.

[Torie]

There are no chops in Marion of townships using the descriptions in Dave's utility. I went by those. But yes, your map has three fewer chops, less the pack penalty, so a chop score better by two - at the cost of a kind of ugly map.

[muon2]

Where DRA uses two letter prefixes for Marion precincts, the townships are accurate. Where there are numbers only, I'm not sure. This is the map.



I couldn't tell from the detail on your map, but the three yellow townships look chopped. I was willing to trade a cover point for 3 county and 3 township chops. My erosity is higher, but with the lower chop count it is Pareto equivalent.

[Torie]

Where DRA uses two letter prefixes for Marion precincts, the townships are accurate. Where there are numbers only, I'm not sure. This is the map.



I couldn't tell from the detail on your map, but the three yellow townships look chopped. I was willing to trade a cover point for 3 county and 3 township chops. My erosity is higher, but with the lower chop count it is Pareto equivalent.

OK.  It is a pain in the butt without Dave utility guidance. The numbered precincts are the old city of Indianapolis, before Mayor Lugar pushed through uni-government. I followed those. I guess they have no meaning anymore, other than the separate precinct numbering system itself. I remember the Wall Street Journal saying the city would then become "forever" Republican. Nothing is forever.

[Torie]

It looks like there are 3 county chops more than the minimum and some number of township chops in Marion. Here's an alternative that only chops Marion, and has no township chops in Marion. This application of the metric gives a 2D, 3r, 4R split with the closest district CD-2 at r+2.

Here is my object d'art which I find visually more appealing than your fine effort, at the cost of one extra chop. Presumably it has a better erosity score to go along with its visual appeal. It has less than a skew than your map, with IN-02 a tossup, and competitive, so that's a two-fer policy wise.

I would never have drawn my first map, if using the old Indianapolis city lines was a no go. Locality chops in a macro-chopped county are almost always a scoring disaster. Your metrics give one a great incentive to avoid that. And presumably that is justified on the theory that it will tend to avoid more substantial partisan variations between maps which are competitive on the pareto optimal frontier.



Do you agree with my road cut count for my artistic endeavor (65 cuts)? Did I do it right? Did I do the count right for the chopped county of Henry? Is there an extra cut when an applicable state highway goes in and then out again of the chop before going into another county (which is what I did)?

[muon2]

By George, I've think you've it.

I just got in, it generally looks good, but I see some missing links. For example, Benton links to White and Tippecanoe. I actually start with a paper Rand McNally map of the state to get links. Then I double check the counties where I didn't see a link on paper. I'll try to put together a more detailed analysis.

In the meantime, I improved my original offering to reduce its erosity. There are still no chops, but it still has the UCC pack penalty. The partisan breakdown has changed to 2D, 1e, 1r, 5R.



Moderator's note: Follow up comments on erosity have been moved to the local vs regional connections thread.

[Torie]

Here's Ohio. Three seat flip to the Democrats - OH-12, OH-14, and OH-01 (OH-01 has a Dem PVI of 1.65%). Thus a Pub skew of 1. Not bad. No tossup CD's alas. The blacks in Cleveland probably would not like the BVAP in OH-13 dropping down to 34.4%, which is probably about 45% of the voters in a Dem primary. If my negotiation regime were adopted, it would be interesting to observe if the blacks pushed the Dems in the legislature to move OH-14 to tossup status anyway, in exchange for getting the Dem black primary percentage comfortably above 50%, by switching around which localities are in which CD.

Notice that I am getting more tolerant of some erosity. I chose to put Noble County in OH-06 to avoid a chop!  

[muon2]

Here's Ohio. Three seat flip to the Democrats - OH-12, OH-14, and OH-01 (OH-01 has a Dem PVI of 1.65%). Thus a Pub skew of 1. Not bad. No tossup CD's alas. The blacks in Cleveland probably would not like the BVAP in OH-13 dropping down to 34.4%, which is probably about 45% of the voters in a Dem primary. If my negotiation regime were adopted, it would be interesting to observe if the blacks pushed the Dems in the legislature to move OH-14 to tossup status anyway, in exchange for getting the Dem black primary percentage comfortably above 50%, by switching around which localities are in which CD.

Notice that I am getting more tolerant of some erosity. I chose to put Noble County in OH-06 to avoid a chop!  

They look quite erose given the number of chops. Am I missing something. I say this because I took a quick glance of maps you helped build in 2013.

[Torie]

OH-06 looks kind of erose as I noted, but otherwise, I don't consider the map particularly erose. It's the balance that I chose. No doubt you can do something better. You usually do.

Back in 2013, we had different metrics. That might explain some of it. Or maybe not. Maybe I am just getting senile.

I do like the way the NE corner of the state is drawn as a package (OH-10-14). I strongly doubt that I would want to depart from that.

Here's a fix for Cuyahoga County. I thought Parma was too big to put in the Cleveland CD, but after the final cut into Cuyahoga of OH-10, Parma worked in OH-13 - barely. So that improves the lines there.



Oh, here is a revised Ohio map that looks prettier.



And here is an alternative map. It has the same chop penalty count as the map above (the chop of Licking is lost, but a pack penalty incurred), but the erosity score is much better. The moral of the story is that I need to get over my pack penalty phobia (PPP syndrome). Avoiding pack penalties will rarely pay if it is at the cost of an additional macrochop. The erosity score will tank. It is sort of like my proposed rule for bridge chops. They won't be worth doing unless it either reduces the chop count, or potentially avoids a locality chop. In that sense, the pack penalty is relatively toothless. It's there, and will make a difference sometimes, but in most instances, only when a macrochop penalty is not in play. And with multi county urban clusters, typically such macrochops will be in play to avoid the pack penalty. Yes, I know, Muon2 figured out that long ago, and keep the secret to himself. But now the secret has been exposed!

The map below also has the virtue of moving OH-8 into more marginal status. It's now an r CD, rather than an R CD. I am also quite sure the erosity penalty is lower when OH-16 is at the northern edge of Franklin. The lines are cleaner. So this map is 4D, 3d, 4r and 5R. I have little doubt that today OH-08 is probably a tossup CD I might add. If so, that gets the skew down to 0.5 Pub. Each side should have 7.5 CD's, but instead the Pubs have 8, and the Dems 7. Maybe Muon2 calculates the skew differently, or numbers it differently, with this map having a skew in 2010 of 2, and a skew of 1 now. I think I like my numbering system better however.