Ohio Redistricting Contest

[muon2]

I just got this congressional plan up.

It's designed to go for the points within the rules, and since there are already two earlier variations on this with strong adherence to county subdivisions, this plan often ignores municipal lines if it would help other aspects of the score.

This plan stuck with the 4 SR, 3 LR, 1 SD, 7 LD, 1 E plan. It has 0.1% electoral disproportionality. To get to 0 required a 3 SR, 5 LR, 2 SD, 6 LD, 0 E plan. CD 12 would have been the best to bring down by running it into Columbus, but the extra split didn't seem worth it for an additional 0.4 points.

All 11 lean districts became very competitive for 33 points in this plan and by carefully tracking the bounding circle limits on district involved in splits the compactness rounds to 49.0%.

It's possible to have fewer county splits than the 9 in a plan, but it reducing splits had a tendency to hurt compactness and prevent maximizing competitiveness.

[jimrtex]

I just got this congressional plan up.

It's designed to go for the points within the rules, and since there are already two earlier variations on this with strong adherence to county subdivisions, this plan often ignores municipal lines if it would help other aspects of the score.

This plan stuck with the 4 SR, 3 LR, 1 SD, 7 LD, 1 E plan. It has 0.1% electoral disproportionality. To get to 0 required a 3 SR, 5 LR, 2 SD, 6 LD, 0 E plan. CD 12 would have been the best to bring down by running it into Columbus, but the extra split didn't seem worth it for an additional 0.4 points.

All 11 lean districts became very competitive for 33 points in this plan and by carefully tracking the bounding circle limits on district involved in splits the compactness rounds to 49.0%.

It's possible to have fewer county splits than the 9 in a plan, but it reducing splits had a tendency to hurt compactness and prevent maximizing competitiveness.

Would a legislature or court sanction including Crestline as part of Richland County based on the 14 persons?  And is it really a good idea to include 30,000 people grabbed out Akron (74% Black and 90% D) to make a Medina district competitive?

Is it possible to get down to 3 Republican sinks or are there not enough Democrats in the right places to match up with them?  Or does the loss in fairness points not worth the extra competitiveness?

[muon2]

I just got this congressional plan up.

It's designed to go for the points within the rules, and since there are already two earlier variations on this with strong adherence to county subdivisions, this plan often ignores municipal lines if it would help other aspects of the score.

This plan stuck with the 4 SR, 3 LR, 1 SD, 7 LD, 1 E plan. It has 0.1% electoral disproportionality. To get to 0 required a 3 SR, 5 LR, 2 SD, 6 LD, 0 E plan. CD 12 would have been the best to bring down by running it into Columbus, but the extra split didn't seem worth it for an additional 0.4 points.

All 11 lean districts became very competitive for 33 points in this plan and by carefully tracking the bounding circle limits on district involved in splits the compactness rounds to 49.0%.

It's possible to have fewer county splits than the 9 in a plan, but it reducing splits had a tendency to hurt compactness and prevent maximizing competitiveness.

Would a legislature or court sanction including Crestline as part of Richland County based on the 14 persons?  And is it really a good idea to include 30,000 people grabbed out Akron (74% Black and 90% D) to make a Medina district competitive?

Is it possible to get down to 3 Republican sinks or are there not enough Democrats in the right places to match up with them?  Or does the loss in fairness points not worth the extra competitiveness?

Let me answer the second part first. I did see that I could extend a tentacle from CD 12 into Columbus and shift it from SR to LR. I could also rearrange CD 9 to swing it from E to LR, but that requires a county split or competitive points lost. Similarly the NE could shift the Akron finger back to 16 or give it to 13 to build an SD out of an LD, but it also shifts a point of competitiveness away from 10. As I noted the 0.4 points were not worth the competitive points lost.

On to the first part, where I really don't care for my plan as policy, but I wanted to illustrate issues in the points system used here, as I went for the max. Here's my policy analysis as sent to the competition. You'll recognize thoughts from our earlier discussion on this thread.

You must be logged in to read this quote.

[jimrtex]

A scoring system for a contest must be carefully balanced to avoid unintended results.

I was looking at some of the other entries and was checking whether the last entry really scored 204.4.  I don't think it does.  But I notice it had 9 Democratic districts and a partisan index of 50.0.

I then started wondering where the 51.4% came from, and found that it is based on the total votes cast for the Republican and Democratic candidates in the 4 races.  Since three races are from 2010 it is more heavily weighted toward 2010, when the Republicans won the three races (Auditor, Secretary of State, and Governor).  Because turnout was heavier in 2008 this effect is somewhat less.   34% of the votes used in the index were from 2008, and 66% from 2010, about a 2:1 ratio.

McCain had 47.7% of the two-party vote in 2008, while the 3 Republicans had 53.4% of the two-party vote in 2010.

2010 thus had a 2.0% R swing relative to the "typical election", while 2008 had a 3.7% D swing.

A district that had been ratcheted in to be a "highly competitive lean Democratic seat (52.4% D), would have been a generally non-competitive strong Democratic district in a presidential election year; and a "highly competitive toss-up, but still slightly Democrat" district in a gubernatorial election.  The complementary district that was a "highly competitive lean Republican seat (52.4% R) would be a "highly competitive lean Democratic district" in a presidential election, and a "generally competitive lean Republican' district during a gubernatorial election.   This assumes uniform swing, which may or may not exist.

[muon2]

A scoring system for a contest must be carefully balanced to avoid unintended results.

I was looking at some of the other entries and was checking whether the last entry really scored 204.4.  I don't think it does.  But I notice it had 9 Democratic districts and a partisan index of 50.0.

I then started wondering where the 51.4% came from, and found that it is based on the total votes cast for the Republican and Democratic candidates in the 4 races.  Since three races are from 2010 it is more heavily weighted toward 2010, when the Republicans won the three races (Auditor, Secretary of State, and Governor).  Because turnout was heavier in 2008 this effect is somewhat less.   34% of the votes used in the index were from 2008, and 66% from 2010, about a 2:1 ratio.

McCain had 47.7% of the two-party vote in 2008, while the 3 Republicans had 53.4% of the two-party vote in 2010.

2010 thus had a 2.0% R swing relative to the "typical election", while 2008 had a 3.7% D swing.

A district that had been ratcheted in to be a "highly competitive lean Democratic seat (52.4% D), would have been a generally non-competitive strong Democratic district in a presidential election year; and a "highly competitive toss-up, but still slightly Democrat" district in a gubernatorial election.  The complementary district that was a "highly competitive lean Republican seat (52.4% R) would be a "highly competitive lean Democratic district" in a presidential election, and a "generally competitive lean Republican' district during a gubernatorial election.   This assumes uniform swing, which may or may not exist.

I haven't seen a master scoresheet, is there one posted somewhere?

I also have wondered about the choice of elections. At one time there was a talk of using the three closest races. That would be AG, Gov, and Auditor from 2010. I'm not sure why they went with SoS instead of AG.

[jimrtex]

2010 thus had a 2.0% R swing relative to the "typical election", while 2008 had a 3.7% D swing.

A district that had been ratcheted in to be a "highly competitive lean Democratic seat (52.4% D), would have been a generally non-competitive strong Democratic district in a presidential election year; and a "highly competitive toss-up, but still slightly Democrat" district in a gubernatorial election.  The complementary district that was a "highly competitive lean Republican seat (52.4% R) would be a "highly competitive lean Democratic district" in a presidential election, and a "generally competitive lean Republican' district during a gubernatorial election.   This assumes uniform swing, which may or may not exist.

I haven't seen a master scoresheet, is there one posted somewhere?

I also have wondered about the choice of elections. At one time there was a talk of using the three closest races. That would be AG, Gov, and Auditor from 2010. I'm not sure why they went with SoS instead of AG.

I have not seen one.  I don't think there ever was one posted after the legislative contest.  There was never a 3rd place winner announced.  So I don't know if they decided because I followed the Ohio Constitution that I couldn't be rewarded.

The AG race was much closer than the SOS race.  The rules say that the 3 offices are on the apportionment board.  But that is really pretty superficial reasoning.  I thought I saw a news article that said that a bill was going to be introduced, but I can't find that any more.  And I didn't see anything on the legislative web site.

I looked up the Democratic candidate in the SOS race, and the only thing that I came across was something about gun rights.  She had been a Columbus city council member and had been filmed in a debate about a law that would have made Ohio's gun carry laws uniform, and had apparently been particularly shrill, and this incident gets remembered a few years later.  So if the Democrats run a particularly inept candidate, and you include twice as many votes from 2010 as 2008, they should get more representatives in Congress?

I was looking through the winning entries, and was looking at the shared version of the 3rd place entry.  Take a look at Lorain County.  I tried to count the splits.  I zoomed in and zoomed in to finally figure out what had been done.

I looked at the two districts, 2 and 13, and thought it pretty weird, but figured maybe they wanted to only get charged 3 points for the triple split of Lorain, vs. the double split of two counties.  But that is really an egregious split, worse than your Akron right hook.

And it was dividing two Republican districts and District 2 included both Richland and Wayne, while District 13 was stretching from Columbus to Elyria.   So I gave all of Lorain to OH-2 and started chopping off bits of Richland, but that was getting into the city of Richland.  So I added some smaller counties to the west (Hancock&Wyandot) to OH-2 and then started taking a few townships at a time from Ashland.  But I couldn't get it right.  One was too high, and then the other.  They were about 5000 apart, and I found a township of about 2500 which would make them both about equal.  They were both too high!

I remembered that the colored map wasn't quite the same.  I thought that the map had been fudged to make the bizarre Lorain crossing less obvious.  But what happened the Toledo district had been underpopulated, and there was apparently a last minute change to get it into range, as well as getting the other two districts below the 0.5% threshold.  So I made the change to Hancock, and was easily able to get OH-2 and OH-13.   I split another county, but at least the districts weren't entangled.

And OH-13 went from 35.2% to 58.3% compactness with its appendages removed, enough to improve the average 1.6 points.  But alas, it had pushed OH-2 to above 57.5% R.  So I had gained 1.6 compactness points and lost 1 competitiveness and added a "split".

I remembered that the colored map had a slightly different Lorain crossing.  So I implemented that,  But it was kind of ugly, so I played around and here is what I came up with.

https://districtbuilder.drawthelineohio.org/districtmapping/plan/1481/view/

This bit of tweaking got OH-2 back below 57.5% R, which gave me a competitiveness point.

But it also pushed OH-11 to a tossup, which changed the partisan index to 52.5% (from 51.2%), which cost 3.6 fairness points.

Since the partisan index is supposed to represent the number of seats one, a 0.3% change means a prediction of 1.3% more Republican seats.  But that 1.3% of 16 seats has to come in OH-11.  Making it 0.3% more Republican is not going to help out in Cincinnati.  1.3% of 16 seats is equivalent to 20.8% of OH-11.  But if 0.3% provides a 20% better chance of getting rid of Dennis Kucinich, would 1.5% make it a certainty?  Clearly not.

So we have a non-representative set of elections, that nonetheless produce a continuous variable.  We convert that into an a pair of arbitrary discrete scores (numerator and denominator) calculate a partisan index which kind of looks like a continuous variable, but you really have to play around with the combinations.

It also turns out that turning the Columbus district into a Strong Democratic seat helps out adding 0.5 to the denominator, is a lot more valuable in terms of "fairness" than the loss of a competitiveness point (it is still under 57.%).

We've created redistricting Yahtzee.

If we wanted to calculate a probable seat yield, wouldn't we just model swing as a normal probability distribution?   In close races, swing may be somewhat uniform - though we do have the local results for the individual races to test that.

BTW, I was playing around with the current map.  I started out merging OH-10 and OH-11 which are the Cleveland seats and the least populous.  This gave an excess for combined district which can be distributed to neighboring districts.  It turns out that the 6 northeastern districts along with the Toledo district have enough population to create 6 districts (7 into 6), which means the other 11 have to be reduced to 10.  The smallest adjacent districts are OH-4 and OH-5.

I was speculating about an incumbent protection procedure where each representative in turn could add to his district.  A district that had 600,000 (121,000) short would add 10% of 12,100 persons.  They would have to take that from one neighboring district with a larger population.  If there was a split county between the districts, then the adjustment would have to attempt to eliminate the split, with similar rules for townships.  A representative who chose later in the same round could not take back the same territory that had just been taken from him.   After all the representatives from districts with less than the ideal population had gone, the representatives from districts with excess population would have to discard some population.  In the case of the two merged districts, each of the two representatives would have to give up some of their former district.

So you would have alternating Push and Pull, or Give and Take rounds.  The percentage could gradually increase as the districts become closer in population.

I couldn't get the representatives to play.  So instead, I assigned each split county to the district which had the largest share,  When there were more than two districts in the 3 largest counties, I assigned the remainder of the county to the district with the 2nd largest share,

I then adjusted the districts based on which county shifts would improve compactness the most (eg it was as if the representatives would always pick a county based on compactness considerations).  The unexpected result was that OH-6 (the Ohio River district didn't have its ends lopped off.  It's compactness is in the low teens.  But it is so long that reducing the  length by one county didn't improve the compactness by that much.  It might be that for that purpose, it would be better to use the reciprocal of the Roeck measurement, or equivalently use the harmonic mean.

ps is Charles related?

[muon2]

So we have a non-representative set of elections, that nonetheless produce a continuous variable.  We convert that into an a pair of arbitrary discrete scores (numerator and denominator) calculate a partisan index which kind of looks like a continuous variable, but you really have to play around with the combinations.

It also turns out that turning the Columbus district into a Strong Democratic seat helps out adding 0.5 to the denominator, is a lot more valuable in terms of "fairness" than the loss of a competitiveness point (it is still under 57.%).

We've created redistricting Yahtzee.

I didn't have it in my most recent commentary, but I noted earlier that the fairness measure is considerably overweighted compared to the other criteria. I would double the disproprtionality rather than quadruple it as in the current rules. That makes the trades for competitiveness and county splits much more interesting. As it stands, if you aren't within a percent on disproportionality it's difficult to get a map score that can compete.

Also, your point about the Columbus district applies if one compares a 54% D district to a 56% D district. I found that having a second strong D district shot one's chance to have 11 highly competitive districts with high fairness unless the pattern was 3 SR - 5 LR - 2 SR - 6 LR.

You must be logged in to read this quote.

He was inspired after the legislative results were announced (and he heard that there was a separate prize for the top HS plan).

[jimrtex]

This is my revision of the official plan.

https://districtbuilder.drawthelineohio.org/districtmapping/plan/1493/view/

I went through the split counties and assigned them to the district with the largest share of the population.  I kept Lucas, Lorain, Medina, Summit, and Montgomery split between two districts.  I made a few adjustments to avoid split districts.

I eliminated the splits of Toledo, Akron, Dayton, and Cincinnati.

I used my version of the Columbus district (OH-3), because it was a bit cleaner.  It might have been modified to keep Upper Arlington out, but that doesn't work very well with a Columbus and south district.  I also used my version of the Cleveland district (OH-11) since it included all of Cleveland and is reasonably VRA compliant.

Because they lost all of Toledo, OH-4 and OH-5 were pushed south but became more visually compact because of it and were able to return Clark to the SE districts.  I rearranged OH-7 and OH-16.  It doesn't make sense to have a district include Lorain and Stark and wrapping around south of Wayne.  The Youngstown-Canton-Akron-Cleveland-Lorain-Toledo area has the population for 6 districts, and one of those should be in the Canton-Youngstown area.  There is room for 5 districts in the more immediate Cleveland area, and we used up our stretch to create the shoreline district.

I also realigned OH-10 and OH-15, so that OH-15 didn't loop around Pickaway and southern Fairfield, through southern Franklin and up through Madison and Union, while OH-15 came across from Dayton to Fairfield.

After putting all of Cincinnati in OH-1 it became very ugly.  It was just barely possible to get up to Butler County when the rest of Hamilton County had to connect with Clermont.  And that still required a split of Butler to make up for the added people from SE Cincinnati.

So I switched to my version of the Cincinnati district which is wholly in Hamilton County.

Because OH-13 was the largest district in Summit, Portage, Trumbull, and Mahoning counties it became very large.  The excess was a consequence of putting all of Toledo in OH-9.  This required shifting Mahoning to OH-6, which in turn moved the southern tip into OH-15 (the Toledo change had pushed OH-4 and OH-5 south through the Columbus area.

I was able to get small count balance through the center of the state, and have 13 split counties.  8 are substantive splits, including the 6 largest counties, Cuyahoga, Franklin, Hamilton, Summit, Montgomery, and Lucas, plus Lorain and Trumbull.  There are population balancing splits in Darke, Hardin. Wayne, Muskingum, and Ross.

I haven't fully balanced OH-3, OH-12, and OH-15, but collectively they are within 476 of the ideal.  The rest can be fully equalized by small adjustments within already split counties:

1 -> 2     913 Hamilton
2 -> 15   372 Ross
12 -> 6  (-103) Muskingum
6 -> 13 (-2045) Trumbull
13 -> 14 1238 Summit
14 -> 11/16 795 Cuyahoga
7 -> 16  (-366) Wayne
11/16 -> 9  327 Cuyahoga
9 -> 5  240 Lucas
4 -> 5 (-177) Hardin
5 -> 8  60 Darke
8 -> 10 (-2534) Montgomery

If you wanted a more favorable Republican plan,

Move Clark to OH-8, and parts of SW Montgomery and N Warren to OH-8, OH-2 takes in more of Hamilton, and OH-1 moves north into Butler County.

You could swap Muskingum for Athens and Morgan, but the structural solution is to put Stark and Mahoning together, which gives you more of a tossup.  The remainder of the OH-6 and OH-7 can be shifted west into Licking, while OH-15 shifts a bit further north along the river.

[jimrtex]

http://drawthelinemidwest.org/ohio/congressionalwinners/

Under "Scoring", "Plan Scores" is a link to an Excel spread sheet with all the scores.

[jimrtex]

This has some links to the congressional and legislative maps from past decades.

http://www.sos.state.oh.us/SOS/ReshapeOhio/Links.aspx

It appears that the constitution has only been loosely followed as to creating legislative districts from whole counties.  I thought we were simply following the constitution, and not trying to eliminate county cuts.

Only in 1980 does it look like there was any effort to combine whole counties.

They have followed the constitution with regard to creating as many districts within the county as possible, for the most part.

It appears that maybe they got into problems in NE Ohio with the house districts though.  Lake County was split into 1 whole district and two parts, and this was also used for the senate districts.

[muon2]

This has some links to the congressional and legislative maps from past decades.

http://www.sos.state.oh.us/SOS/ReshapeOhio/Links.aspx

It appears that the constitution has only been loosely followed as to creating legislative districts from whole counties.  I thought we were simply following the constitution, and not trying to eliminate county cuts.

Only in 1980 does it look like there was any effort to combine whole counties.

They have followed the constitution with regard to creating as many districts within the county as possible, for the most part.

It appears that maybe they got into problems in NE Ohio with the house districts though.  Lake County was split into 1 whole district and two parts, and this was also used for the senate districts.

On your last comment, I assume you are referring to the 1970s map. Indeed Lake did not have enough for two full house districts and the population in NE OH also seems to force at least one unconstitutional construction. The treatment of Lake seems like my split of Cuyahoga, though I see that Mahoning also does not get a constitutional senate district in the 1970s plan.

[jimrtex]

This has some links to the congressional and legislative maps from past decades.

http://www.sos.state.oh.us/SOS/ReshapeOhio/Links.aspx

It appears that the constitution has only been loosely followed as to creating legislative districts from whole counties.  I thought we were simply following the constitution, and not trying to eliminate county cuts.

Only in 1980 does it look like there was any effort to combine whole counties.

They have followed the constitution with regard to creating as many districts within the county as possible, for the most part.

It appears that maybe they got into problems in NE Ohio with the house districts though.  Lake County was split into 1 whole district and two parts, and this was also used for the senate districts.

On your last comment, I assume you are referring to the 1970s map. Indeed Lake did not have enough for two full house districts and the population in NE OH also seems to force at least one unconstitutional construction. The treatment of Lake seems like my split of Cuyahoga, though I see that Mahoning also does not get a constitutional senate district in the 1970s plan.

Yes.  I was referring to the 1970s map.

The constitution was changed in 1967, so I think that the the 1972 plan was the first to use the new system.  I find it interesting that there was no real attempt to form districts from whole counties, but then in 1982 there was.

[jimrtex]

Democrats have filed a lawsuit against the Ohio legislative districts claiming that it violated the Constitution, and asking that new districts be put in place before the March primary.

Ohio Supreme Court]Ohio Supreme Court Case Search, search for 2012-0019

The complaint muddles up the legislative and congressional redistricting which were done by two different bodies.

It also misinterprets 11.07(c) which says that only one geographical unit (township, city, ward, etc.) may be split between two districts.  The erroneous interpretation is that a single district can only contain parts of one geographical unit.  But a district does not divide a geographical unit; but rather a geographical unit is divided between two districts (or among 3 or more districts).

They might have a better case based on 11.07(a) and 11.07(b), which require a district to be comprised of whole counties or whole geographical units.  But this has not been rigorously enforced in the past, and it is unclear how to apply this to single districts, rather than the totality of the map.