US House Redistricting: Michigan

[dpmapper]

I redid the non-Detroit parts a bit:



I used CD-03 to give me another county split involving the Flint district, rather than soak up Battle Creek/Albion.  This lets Flint take Saginaw.  So CD-07 no longer looks like a game of Jenga, and no longer has to run into Saginaw, but it does have to move about 20K farther in towards Ann Arbor (there are some 65%+ Kerry precincts on the outskirts) and take back Battle Creek/Albion (and Mark Schauer).  It does get heavily GOP Barry County in the bargain; I haven't totalled everything but I think this makes it almost 55% McCainBush, compared to 53.7% previously.  

I think Camp will be a bit safer as well - his district goes up to southern Grand Traverse County and takes in less of Saginaw's Dem suburbs than it used to.  

The CD-08/CD-09 combo that splits Oakland takes in a bit more of the Lansing area, but is compensated by the black AnnArbor/Oakland/West Detroit district eating up Ferndale and a bit more of Farmington Hills.  One will have to juggle the split of Oakland carefully to make sure both districts are reasonably solid; I haven't done the number crunching yet.  

[In all of this, I'm assuming that the county splits metric is changed to something more sensible.]

[Torie]

In other news, I sweet talked this really nice lady at the Wayne County elections department to fax me all 44 pages of Wayne County precinct returns for 2008 at no charge. She was particular impressed that I could name all 40+ towns in Wayne from memory. She could just sense how much I cared. And I shared with her the story of Inkster just for bonding purposes. I am getting really dangerous out there! 

[ilikeverin]

You seem to have a spare Okemos precinct in the Lansing/Flint district.  Have you split it Okemos/East Lansing, then?  That would be... strange, but Okemos would probably like it.

[dpmapper]

You seem to have a spare Okemos precinct in the Lansing/Flint district.  Have you split it Okemos/East Lansing, then?  That would be... strange, but Okemos would probably like it.

Yeah, I know.  It's a precinct that has two disjoint parts, one of which is part of the East Lansing block.  I don't know what the rule is about those. 

[Torie]

You seem to have a spare Okemos precinct in the Lansing/Flint district.  Have you split it Okemos/East Lansing, then?  That would be... strange, but Okemos would probably like it.

Yeah, I know.  It's a precinct that has two disjoint parts, one of which is part of the East Lansing block.  I don't know what the rule is about those. 

That is a really hideous looking map! 

[dpmapper]

That is a really hideous looking map! 

Um... thanks?

[TheDeadFlagBlues]

The rough draft of my Michigan gerrymander:




I tried to eliminate Levin instead of Peters and I think I did a fairly good job of it. I packed the yuppie/hipster portions of Oakland County into Conyer's district along with Pontiac, which would significantly weaken both Peters and Levin, who would probably try to attempt a kamikaze run against Candice Miller. Unfortunately even with the placement of the Black parts of Macomb County into Clarke's district, Miller's district is only R+1. As for the rest of the map, it's pretty simple. I packed Lansing into MI-5 and tried to strengthen Benishek and Walberg somewhat. Rogers got an ultra-GOP seat thanks to Miller's district being totally dismantled along with Levin's. Everything else is the same. 

[I spent the winter writing songs about getting better]

The way Oakland County is drawn would definitely be illegal.

[muon2]

I decided to follow up on my earlier observation that two black-majority seats could be drawn in Wayne. To that map I applied the definition of county splits as applied in 2001 to draw a GOP-oriented map that achieves the ideal number of 9 county splits. I did not use incumbent data, and estimated partisan strength based on previous posts and Atlas maps.

Placing two black-majority seats (CD 13, 14) in Wayne leaves 3 fragments: in the Pointes, Livonia, and the south suburbs. To keep Wayne to one split requires that two be joined together. The political goal requires that Ann Arbor and the southern end of Oakland be drawn together in CD 11. These were linked through the Livonia fragment and then across southern Macomb to the Pointes fragment.

Enough population was left to create a whole CD in each of Oakland (9) and Macomb (12). To minimize county splits remainders of those two counties had to be linked to a whole number of counties. In this case CD 10 uses a small part of Macomb to achieve population equality. CD 5 links all of Genesee to Pontiac and Auburn Hills in Oakland.



The rest of the state is completed with a minimum number of splits. Of note is pulling East Lansing out of CD 8 which is stuck with Ypsilanti. East Lansing and Saginaw are now both in CD 4.



For those of a mathematical bent, the county minimization was solved with graph theory. Each node is a district and each link is a split county. There are 10 nodes since 4 districts are entirely within one county. The minimum of 9 links occurs when the graph has no loops. The graph for my map is the following.

            [1]                [5]        [10]
            /                             /
           /                             /
        [2]----[4]----[8]----[11]
          |                           |
          |                           |
        [6]         [3]             [7]

[dpmapper]

Muon, making the graph of county splits a tree is exactly what I was going for with my metric for county splits.  I didn't state it in that language because I wasn't sure anyone would understand me.   

But that's about the only map I've seen that is a plausible gerrymander that nevertheless minimizes county splits both by my (sensible) standard and the 2001 standard simultaneously.  Nicely imagined. 

It seems that districts 12, 7, 8 might be a bit too marginal for the GOP's taste.  I think Rogers's district gaining not only Ypsilanti but also the Ann Arbor outskirts like Pittsfield and Scio Townships will outweigh losing East Lansing.  7 might be fine if not for its northern-most three cities.  And McCotter is kind of screwed. 

[Torie]

MI-07, MI-08 and MI-12 are all marginal, with MI-08 looking even lean Dem in Muon2's map.  If that is what Michigan law dictates, the law is going to be changed. What I am trying to figure out is whether the minimum number of splits were done in 2000.  I don't think that was the case.

[dpmapper]

I've finished tabulating the numbers on the districts in my latest map.  

McCotter (light green) is still at 53.06% Bush, up slightly from 52.6% currently.  Not much can be done here, the way I have his district set up, but the good news is that voters here have probably been trending GOP the last 6 years.  

The two GOP districts that split Oakland are at a combined 55.5% Bush.  I didn't bother to calculate the intra-Oakland numbers but there must be a way to split it so that both stay at or above 55% Bush.  

Walberg's 7th (in grey) is at 54.67% Bush.  A bit lower than one would like it, and a bit lower than it is currently (54.97%).  Some of the areas outside Ann Arbor really hurt it.  With that area booming (by Michigan standards) this might be cutting it a little close.  Can't do much to help that without upsetting the entire apple cart.  

Upton's 6th (teal) is over 56% Bush.  Some almost-90% GOP townships in eastern Ottawa really work wonders.  

Camp's 4th (red) is at 54.90% Bush.  His part of Saginaw County actually helps matters, as it's at 56.5% Bush.  

Benishek's 1st (blue) is at 55.4% Bush.  

2nd (green) and 3rd (purple) will be quite safe, as even before adding in Kent County numbers, the 2nd is at 55.8% Bush and the 3rd is at 59.6%.  I won't bother with running the intra-Kent numbers; the mapmakers should feel free to split Kent in whatever way is most logical, rather than worrying about partisan numbers, as long as you put all of the minority parts of Grand Rapids into the 3rd (which would be the logical thing anyway).  

Haven't looked at Miller's numbers but she should be comfortable.  Her current district was over-Pubbied to get rid of Bonior.  

[muon2]

MI-07, MI-08 and MI-12 are all marginal, with MI-08 looking even lean Dem in Muon2's map.  If that is what Michigan law dictates, the law is going to be changed. What I am trying to figure out is whether the minimum number of splits were done in 2000.  I don't think that was the case.

The 2000 map has the minimum number of splits given only three districts are wholly inside a single county. I see that the 2000 map did not place a district wholly inside Macomb as I did. If I assume that they did minimize county splits according to some rule, and did not make a Macomb district as I did, then I would have to revise my assessment of their counting rules.

I would now conclude that if two discontiguous parts of a county are attached to other counties, but not to the same district then that counts as two county breaks. That is still consistent with the current map in Oakland where one split is divided between CDs 8 and 11, and a separate split has the piece of CD 12.

With this interpretation, which is the only way to legally justify the 2000 map, my map would now count as 11 splits not 9. To get to 10 splits I need to have the pieces of CDs 10 and 11 adjoin in Macomb. That would require running a thin line of CD 11 across Warren and another thin line up to New Baltimore. Since that would split two towns it fails. I suspect that is why a Macomb-only district didn't appear in 2000. Even so it still leaves my map with the two-split of Wayne that cannot be eliminated if two majority-black districts are both in Wayne.

[Torie]

MI-07, MI-08 and MI-12 are all marginal, with MI-08 looking even lean Dem in Muon2's map.  If that is what Michigan law dictates, the law is going to be changed. What I am trying to figure out is whether the minimum number of splits were done in 2000.  I don't think that was the case.

The 2000 map has the minimum number of splits given only three districts are wholly inside a single county. I see that the 2000 map did not place a district wholly inside Macomb as I did. If I assume that they did minimize county splits according to some rule, and did not make a Macomb district as I did, then I would have to revise my assessment of their counting rules.

I would now conclude that if two discontiguous parts of a county are attached to other counties, but not to the same district then that counts as two county breaks. That is still consistent with the current map in Oakland where one split is divided between CDs 8 and 11, and a separate split has the piece of CD 12.

With this interpretation, which is the only way to legally justify the 2000 map, my map would now count as 11 splits not 9. To get to 10 splits I need to have the pieces of CDs 10 and 11 adjoin in Macomb. That would require running a thin line of CD 11 across Warren and another thin line up to New Baltimore. Since that would split two towns it fails. I suspect that is why a Macomb-only district didn't appear in 2000. Even so it still leaves my map with the two-split of Wayne that cannot be eliminated if two majority-black districts are both in Wayne.

As I see it, you get one free chop depending on what county you start counting from. That principle derives from the concept that if you have two CD's in 3 counties, with each CD taking one whole county, the split of the third county counts as but one chop. So in your map, either you count the MI-07 salient into Wayne as a chop, or the MI-10 salient into Macomb as one. Per this way of counting,  you have 5 chops for 8 CD's (MI 5, 7, 9, 10, 12, 11, 13 and 14). If you choose to count the MI-10 chop into Macomb as a chop, then from Wayne you have one chop into Oakland with MI-11, another chop into Oakland with MI-05, a third chop into Macomb with MI-11, a 4th chop into Macomb with MI-10, and a 5th chop into Wastenaw with MI-05.  Your free chop is MI-07 in to Wayne,. Since you avoid chops on the other side of MI-05, MI-10 and MI-07,  the only population equalizing mechanism out of the geographic area encompassing the 8 CD's is via Washtenaw and MI-05. That is about as efficient as you can get I think. You get 8 CD's covered with only 5 chops, for an efficiency rating of 3.

I am trying to get the same efficiency rating of 3 with a map that might be acceptable to the Pubbies. As I said, your map just isn't acceptable from a Pubbie perspective. (The first and most immediate thing to do is get MI-13 into Macomb, and MI-11 the hell out; another I think is to get MI-05 to suck up both Pontiac and Ann Arbor.) Either there is an alternative map that works from a partisan perspective, or there isn't. If there isn't, then we know that the law will have to be changed.

This is a tough nut to both understand and then crack. It's a nightmare. I also suspect that the 2000 map does have a maximum efficiency rating, but I am still not entirely sure.

[Torie]

Well here is my version of a plus 3 minimum chop map. Yes, MI-11 is short by 3,200 or so folks, offset (deliberately) by an equal excess in MI-05, but that can be equalized by shaving the size of MI-05's two link precincts to Pontiac in far north West Bloomfield, down from 5,500 or so in population down to an equalizing number of around 2,300.  



No precinct can be switched there, because that would mean that MI-05 is bounced entirely from Pontiac (or mostly), and the whole map collapses. Making the city chops "work" in Wayne was also a rather terrifying exercise, but I found the slot to squeeze between the Scylla and Charybdis legal monsters to make that work too - again just barely.

At least while the strong Pubbie incumbents hang around, this map actually has a lot of potential if, and only if, the Pubbies have guts - a lot of guts. McCotter in MI-11 has but a lean GOP seat I suspect, but can probably hold it, while Dingell is axed. And Candice Miller should be able to hold MI-12 if she has guts. MI-09 is also probably a lean GOP CD (maybe even in the marginal category, albeit on the GOP side, with a PVI as low as +1% GOP PVI potentially), considerably short of "safe" in any event, so that should be a barn burner of a contest for the incumbent Dem Peters, (but the Pubbie should have a edge I would think in 2012 at least, with economic issues dominating the landscape almost totally is my guess). It will be a close contest however. Overall, if all goes well, and assuming the Pubs choose a competent nominee to run not only in MI-09 but also in weak safe to lean GOP MI-10, and assuming further that Walberg can hold a weak safe to strong lean GOP CD in MI-07, that means the Dems will be held to but 3 seats in Michigan. (Rogers should be able to hold his CD from his Livingston base despite having to such up half of Ann Arbor and all of Ypsilianti in Wastenaw County (both 3-1 Obama - I gave Rogers the more GOP half of Ann Arbor -the balance was more like 85-15 Obama ), with the Dutch via Kent and Ottawa being used to pacify Lansing.) MI-02 and MI-03 will both become considerably weaker "safe" CD's for the GOP.

Everything was just so close in this map, and it threatened to collapse at any moment. Notice how both black CD's are just barely over 50% black VAP. And there was no margin for error at all vis a vis MI-05. I lived in terror of its Pontiac salient collapsing. And then I need to play the county game with MI-10, and MI-07 to get the numbers very close to what was required for a Detroit metro map where there was very little play at all - almost none. I had maybe 20,000 folks to pay with. Yikes! But there was, just barely, an objective function which emerged for the Pubbies, at least for those with some courage.

Will the Pubbies have such guts? I tend to doubt it. They will probably change the law. Among other things, the Pubbie incumbents will perforce be saddled with a lot of new territory, in the case of most of them. But it should be an interesting discussion behind closed doors!  They should hire me as a "consultant" to "help" them with some of this.  

By the way, a note of caution to you Michigan cartographers. Some of what look like "cities" in the Michigan map as depicted by the Dave Bradlee software are but villages, and villages under the Michigan law don't count as equal to townships from it comes to intra-county splits - only cities do. So you can't split a township claiming you are just sucking up a village; that dog just won't hunt legally. For example, all those little areas with lines around them in Southfield Township north of the City of Southfield, are but villages, and so they must all be in but one CD, unless your chop is going to be in Southfield Township rather than somewhere else.

[Torie]

And here is an all Oakland use for MI-05 that might actually be a bit safer, at least for McCotter in MI-11.  He could be made even safer with a more favorable Washtenaw chop, but that would be at the expense of Rogers in MI-08.

[Sbane]

How would that 9th vote? About d+2 or 3?

[Torie]

How would that 9th vote? About d+2 or 3?

Yes, I don't think the numbers will look very good. It could be that Dem. It certainly is no better for the Pubbies than marginal. As I said, the Pubbies if they want to ax Peters (and make maybe as many as 4 other Pubbies reasonably safe, rather than on the edge), will need to change the law, or get a friendly Michigan Supreme Court. They should be nervous.

In fact, the numbers really suck for MI-09.



So the map fails. So it is either Muon2's map, with MI-11 becoming the new Dingell CD, or for some Dem, with McCotter having to move to Oakland, or the law will need to be "clarified,"  with some "helpful" amendments thereto. The problem with Muon2's map however is that extra chop into the Pointes by his wandering MI-11 CD, so it fails the efficiency test, since with one less chop per my map above, we can still have two 50% black VAP CD's. Thus, the Pubbies can't hide behind the VRA as hoped. I don't think there is an objective function here that works for the Pubbies.

Oh wait, we decided that two discontiguous chops by one CD into the same county counted for only one chop, so I guess that Pointes salient does not count as another chop. So we will have to see what we can do, if we clean up Muon2's map a bit. But not today. But it might be something like the map below.

Addendum: In fact my latest iteration of the Muon2 map (with MI-14 taking those Macomb precincts rather than MI-11) should work quite well for the Pubbies actually, with the possible exception of Walberg in MI-07, who may have a marginal seat with all that Wayne territory with which he is saddled in this map. However, his string of counties along the southern border of Michigan are quite heavily Pubbie. I don't think there is another map within existing law that works nearly as well. The advantage of this map, is that the cordon sanitaire in Washtenaw has been extended to about just the right place to keep all those massively Dem precincts in Washtenaw nicely contained on the opposite side of the Pubbie acceptable zone. Livingston County just can't handle Ann Arbor and Ypsilanti at all. The Dems have a 100,000 vote margin there.

Avoiding an illegal double split of townships or cities in Oakland while keeping MI-09 wholly contained in Oakland with just the right population with this map was quite a challenge. But I found a way!  

[Torie]

Here are the stats for the new MI-07 per the new map. The seat was marginal, and remains marginal, but with more of a Dem list. I am more convinced than ever, that if the Pubbies don't try to change the law, this is the map. Walberg's weakness in MI-07 will be the gain for Rogers in MI-08 and Camp in MI-04 in particular, both far more powerful and important Pubbie politicians.

The PVI's in a lot of Pubbie seats in Michigan in any event are going to look rather anemic per 2008 figures (the trend in MI from 2004 to 2008 must have been a couple of points to the Dems), absent doing a full throated Gerrymander. (See my Pennsylvania and Ohio and Indiana maps for what a "just go for it" Gerrymander looks like; I was a bit more restrained with Wisconsin for some reason, probably because more erosity would not change the ultimate outcome much.)

[Torie]

I added MI-04 to the list of CD's which feed through the Wastenaw MI-11 population equalizing spigot, in an attempt to get the Pubbie numbers up a bit in MI-12, MI-10, and MI-07, all of which per the map above were close to dead even in a PVI sense. I must say, MI-14 looks ever more interesting as a CD; it is getting to be quite a work of art!  I also manged to get Garden City into MI-07 (labeled MI-08 in the map below; I will change the number for next time to the proper one).  

[Verily]

You appear to have cut Center Line off from the rest of MI-12 on that map, Torie.

[Torie]

You appear to have cut Center Line off from the rest of MI-12 on that map, Torie.

 

Appearances can be deceiving. I do make mistakes in maps, but that is not one of them.   Thanks however for the comment, and anytime you think you see a flaw, I would appreciate your letting me know. I must say, that at this point, I think I know almost every Warren precinct by heart now, as the precincts go into MI-13, and then out again, and then in, and then out, and on and on.

[Torie]

Here is the tentative map. I think I broke the code about county splits per Michigan law. You start with one clean line, and from the opposite side of the state, limit one CD to just one chop.

In 2000, the clean line was the north end of MI-10, and the one chop CD was MI-01. In my map, the clean line is the east end of MI-07, and either MI-06 or MI-02 will have but one chop. Actually, MI-06 has no chops, but that is an accident. There will be a tiny chop. Whether MI-06 loses its excess of 700 folks by MI-02 dipping down into Allegan, or MI-03 doing the dip, is a choice, and an irrelevant one. So we have the same number of chops as the 2000 maps, less one for one fewer CD, but plus one, because the VRA requires an extra chop involving one of the two black CD's in Wayne, with either MI-12 taking the Pointes in Wayne, or one of the black CD's chopping into Macomb. I am very confident that this map is legal.

That aside, I think the partisan numbers will work pretty well, with a lot of weak safe to lean GOP CD's, with maybe MI-07 slipping into the the marginal category, with a GOP lean.

My main concern is MI-03. It may have been put under too much stress, taking in both Lansing, and losing a lot of Kent County suburbs, although the big one MI-02 took is probably fairly Dem per the demographics. The pawns may have to be shifted around in Kent. It may be that the black precincts in the city of Grand Rapids will have to be split up between MI-02 and MI-03. And maybe MI-04 will have to drop Ionia, and be jiggled around. That is the potential weak spot in the map - the issue being just how to most effectively neutralize Lansing - to the extent it can be.

Oh yes, I know that one CD is off by about 3,000 folks, but I have no intention of playing with the other 13 less than 1,000 in population off CD's to deal with that.

[Verily]

What's the PVI on that Grand Rapids-Lansing seat? It can't be better for the GOP than marginal.

[Torie]

What's the PVI on that Grand Rapids-Lansing seat? It can't be better for the GOP than marginal.

I don't know; that was the concern that I expressed in my post above.  You may well  be right. But in the end, both MI-02 and MI-03 I suspect can be made weak safe to strong lean GOP CD's - with a PVI of from say 2.5 to 3.5. It just requires the right chop of Kent, and maybe weakening MI-04 a bit. Yes, the map is on the cusp. The GOP may just dump the law. A lot of efficiency is lost by putting all those Pubbie link townships in Oakland in MI-05, and MI-11 sucking up GOP friendly Livonia, and Northville, and Plymouth in Wayne. That is what the law does, if given a tight interpretation, with not much play afforded in what is deemed "reasonable," a term that is used in the law when it comes to chops.

The big winner in playing this legal game, is the Pubbie in MI-01. That dude will be thrilled!    It was due in part to the imperative of MI-10 having clean lines on its north end. MI-01 was drawn almost in its entirely to accommodate other considerations.