Washington Legislative Districts

[muon2]

A fairer test of how the political balance should go requires drawing the districts from scratch according to a neutral set of rules.

Before I would draw districts, I first draw up regions of whole counties that contain whole numbers of legislative districts such that they can be drawn within 5% of the quota. This minimizes the number of split counties.

Here is an example with the number of legislative districts in each region. The representative districts can then be drawn in each region.

[muon2]

As I noted ealier, I wanted a baseline to compare the current WA assembly with the potential advantage of split districts. The starting point was to draw revised legislative districts according to neutral rules. I started with this set of whole county regions with each region containing a whole number of districts. Districts here are required to be within 5% of the population quota, though the actual districts are much closer in population.



Then I divided each of the regions again allowing a 5% deviation from the quota. Divisions were made to minimize county chops, reservation chops, city chops within a county, and school district chops within a county in that order of priority. There's one extra county chop in Lewis county due to the geography of Fort Lewis going to the shore. That resulted in this plan.




The political division of the LDs is
21 uncompetitive D (PVI 6+)
5 competitive d (PVI 2-5)
8 even (PVI 0 or 1)
2 competitive r  (PVI 2-5)
13 uncompetitive R (PVI 6+)

This is a SKEW of 1 for the Dems, that is they only have one more seat (or the Pubs one less) than the statewide vote from presidential years would predict. In 2012 the national vote was D+2 so one would expect the even seats to swing Dem resulting in a House with 68 D - 30 R, which clearly isn't reached because of the effects of incumbency. In 2014 the national vote was R+3 so the even districts would be expected to swing Pub as would one or two of the lean d seats. Without incumbency the House might be expected to be 49 D - 49 R.

With this in hand I can now look at how the same map would perform if each seat is split into two using the same rules.

[muon2]

To complete the test of nesting districts on the WA partisan balance I created a plan of 98 House districts based on the 49 LDs in my previous post (similar colors show the nested pairs).



The political division of the HDs is
39 uncompetitive D (PVI 6+)
17 competitive d (PVI 2-5)
12 even (PVI 0 or 1)
9 competitive r  (PVI 2-5)
21 uncompetitive R (PVI 6+)

This is a SKEW of 6 for the Dems. The LD map with 49 districts was 1 Dem and for the House with 2 seats per LD the SKEW would be 2 Dem. This nesting of the LDs has tilted the balance in an even year towards the Dems.

In 2012 the national vote was D+2 so one would expect the even seats to swing Dem resulting in a House with 68 D - 30 R, which is the same as would be predicted with 2 elected from each LD. In 2014 the national vote was R+3 so the even districts would be expected to swing Pub as would about third of the lean d seats. In that case House might be expected to be 50 D - 48 R, which is a change of only one seat compared to the situation with 2 per LD.

The change is more noticeable in the SKEW than in the actual House composition. This is because the biggest change between the two plans was the increase in seats that lean for one party (PVI 2-5). The Pub lean seats were matched by a reduction in solid seats. But the Dems lean seats came more at the expense of the even seats than the solid seats.

Looking at the detail by region one can see where nesting has the most effect. In King county there are 14 LDs and they went 13D-1e. When they split it became 23D-4d-1r, where the even LD split into a lean d and lean r HD and three other solid LD created 1 lean d seat each when they split. But that's no change in the SKEW.

In the northern Puget Sound region there are 9 LDs that went 4D-1d-3e-1r, and the split into 18 HDs went 8D-4d-3e-2r-1R. This increased 1 Dem SKEW, but the biggest change was that even seats were reduced creating more lean seats. In the southern Puget Sound region there are 10 LDs that went 3D-2d-2e-1r-2R, and the split into 20 HDs went 6D-4d-4e-3r-3R which is essentially no partisan change. Overall nesting had little effect on the partisan makeup of seats from the greater Puget Sound area.

In the southwest and west regions there are 5 LDs that went 1D-2d-2R, and the split took it to 1D-4d-2e-2r-1R. This is noticeable change with the more polarized LDs adding many more lean and even HDs. The SKEW however didn't change since it there was 1 excess Dem LD and 2 excess Dem HDs which is expected from a split into two parts.

In the next post I'll show how the major changes happened in eastern WA.

[muon2]

There's a trade off involved with nested districts. If I form regions by the Senate district I minimize the county chops at that level, but there may be more chops (particularly of subunits) to form House districts from them. If I form House district regions first then wherever there is an odd number of HDs in a region I'm likely to end up with an LD that spans county lines, that is I tend to increase county chops at that level. The best way to avoid chops is to de-nest the HDs from the LDs.

As an aside, the Whitman-Asotin HD only exists under the rules by forming the LD first. Asotin has no state highway connection to Colfax in Whitman that doesn't leave those counties so I couldn't for it directly. However, there is no way to divide the larger LD without using at least link involving local roads, so the version that keeps the counties whole is preferred.

The only reason the Snohomish district dips into King is due to the rule that in a macrochopped county (like King) all the subunits have to be connected. The Skykomish school district along the Stevens Pass is isolated from the rest of King. Its population is so small it could be added to the district to its west, but the rules require it to connect to Snohomish instead.

[muon2]

Splitting the LDs into HDs had a substantial effect in eastern WA. In the LD plan that set of regions had a political breakdown of 2e-9R. There were no districts that favored Dems at all. In the HD plan the same area was 1D-1d-3e-1r-16R. Not only are there 2 districts now that favor Dems, but one is solid D. In term of SKEW the net effect was to shift 3 points to the Dems.

In most of the regions of eastern WA there was no effect. Both parts of the Ellensburg-Moses Lake LD are just a solid R as the whole. Same thing with all four parts of the 2 LDs in the Tri-Cities area where the Hispanic population doesn't (or can't) vote and the rest is very Pub. The one lean r HD comes from the effect of Washington State Univ in Whitman county.

The shift in the distribution comes from the two large cities in the east: Spokane and Yakima. Spokane county has a population for about 3 1/2 LDs or 7 HDs. The city itself has more population than 1 LD, so it was divided along the Spokane river. The part of the city north of the river was one LD (reds) and the part south plus the west and south school districts in the county formed the other LD (cyans). Both of these were politically even with PVIs of 0 or 1. However when they were split the light red is a lean D (D+4) and the light cyan is a solid D district (D+9), leaving the dark red HD even (D+0.1) and the dark cyan HD as solid R. It is interesting to note that adding the reds together only creates a D+1 LD, but that's because the dark red provides 50% more votes than the other half and dominates the result.



The other significant change in skew comes from Yakima. Both the salmon and green combined LDs are solid R. However, the light salmon HD in the Yakima school district part of the city is an even political district, though it has half the voter turnout of the West Valley school district part of the city due to the high Hispanic population in the east part of the city. The green HD north in NE Yakima county has an even higher Hispanic population (HVAP 52%) than the Yakima city HD (HVAP 42%), but lacks the urban white Dem crossover vote of the city, so the green HD stays solid R. The light green HD in south Yakima county includes Klickitas and Skamania and is 31% HVAP, but also is 9% NVAP from the Yakama reservation. Those coalition minority votes are enough to make it an even political HD.

[muon2]

No UCC areas for this exercise?

For this exercise I just concentrated on county and subunit integrity. King came out to 14.07 LDs, so I placed it in its own region and worked from there. The only UCC violation is from that choice and its impact on the Seattle and adjacent UCCs.

Since the interesting results were in eastern WA, I don't know that the UCC factor would make a difference in the conclusions about nesting. There are four yellow-green counties on jimrtex's WA UCC map, but none are in a UCC (I'm guessing they are left over from an earlier map) so there are no violations in the east. However the three county Seattle UCC is a nice 25.07 LDs so I will take a look. In the meantime my clean connection map examples for the other thread will have to wait.

[muon2]

I found a plan of LD regions that as many as my previous plan and keeps the Seattle UCC as a single region. The Olympia region is 11,191 under the quota for three LDs, but this is less than 5% per district. It's just within the limit of 5%*sqrt(N) for N districts as suggested once by jimrtex as a guide for maximum regional deviation.



I used this to rework the LDs in western WA, then draw HDs from those LDs. Starting from this regional plan meets all UCC packs and covers, and reduces the number of county chops at the increase of subunit chops. Some may also appreciate that I as able to attach Island to Skagit. Here's the greater Seattle-Tacoma area. As before, similar colors (light and dark) represent the pairs of HD in an LD.



At the LD level the plan is 21D-6d-7e-2r-13R, which shifts one seat from even to lean d for a SKEW of 2D. At the HD the plan is 39D-19d-8e-11r-21R, which shifts 2 even HDs to lean d and 2 even HDs to lean r, thus keeping the SKEW at 6D.

Effectively there is little partisan difference in SKEW by keeping the Seattle UCC intact and a modest increase in POLARIZATION due to the loss of even seats. That increased POLARIZATION suggests that there would be less swing in seats between elections such as 2012 to 2014. Nonetheless the main conclusion remains that dividing LDs into HDs helps Dems due to the remote urban centers in eastern WA, and changes little in the dense urban regions of western WA.

[muon2]

I found a plan of LD regions that as many as my previous plan and keeps the Seattle UCC as a single region. The Olympia region is 11,191 under the quota for three LDs, but this is less than 5% per district. It's just within the limit of 5%*sqrt(N) for N districts as suggested once by jimrtex as a guide for maximum regional deviation.



I used this to rework the LDs in western WA, then draw HDs from those LDs. Starting from this regional plan meets all UCC packs and covers, and reduces the number of county chops at the increase of subunit chops. Some may also appreciate that I as able to attach Island to Skagit. Here's the greater Seattle-Tacoma area. As before, similar colors (light and dark) represent the pairs of HD in an LD.



At the LD level the plan is 21D-6d-7e-2r-13R, which shifts one seat from even to lean d for a SKEW of 2D. At the HD the plan is 39D-19d-8e-11r-21R, which shifts 2 even HDs to lean d and 2 even HDs to lean r, thus keeping the SKEW at 6D.

Effectively there is little partisan difference in SKEW by keeping the Seattle UCC intact and a modest increase in POLARIZATION due to the loss of even seats. That increased POLARIZATION suggests that there would be less swing in seats between elections such as 2012 to 2014. Nonetheless the main conclusion remains that dividing LDs into HDs helps Dems due to the remote urban centers in eastern WA, and changes little in the dense urban regions of western WA.

Would it be feasible to go from Grays Harbor to Lewis, avoiding the double chop of Thurston?

Also, why not put all of Jefferson in the same house district?

Are the only districts that cross the Pierce-King and Snohomish-King line the eastern districts (light brown and aqua)?

Good questions.

There's no connection from Grays Harbor to Lewis so the double chop of Thurston was unavoidable. As I noted, I had missed that three county combination originally since I saw the two county Grays Harbor Mason combination for one LD. When pushed to keep the Seattle UCC intact it appeared as an option.

I should put Jefferson all in one HD. I was used to keeping all parts connected, but we are now working under the rule that all parts are connected if whole. I'll change it in my master version, but probably not update the maps here unless there are enough changes that it might affect the SKEW computations, since that is the purpose of the exercise.

Yes I used the minimum number of county crossings from King to its two UCC neighbors. There are subunit chops in those HDs.

[muon2]

Joining Jefferson and Island Counties is a no-no.

Eh, there's a regular ferry between Port Townsend and Coupeville/Keystone.

I'd say it fits much better with Skagit, since Camano Island connects to Stanwood and Whidbey Island to Anacortes, but Jefferson isn't indefensible.

I'm from Clallam county, so consider me the forum's Olympic Peninsula expert. The Coupeville ferry exists, but it's very small, not always active (Lot of choppy weather in that stretch), and rarely used. Plus I consider the number 1 rule of Washington map making to be "Don't cross the Sound"

As for chopping Jefferson county, it's worth noting that the western 3/4 of Jefferson are mostly national park land and some sparsely populated areas. Those areas have more in common with rural Clallam or Grays Harbor county than the Port Townsend half of Jefferson.

I'm surprised at your description of the ferry. I used it this last summer traveling from Bellingham to Sequim and it was extremely busy with a long line of cars and ferries every 45 minutes.

[muon2]

Splitting the LDs into HDs had a substantial effect in eastern WA. In the LD plan that set of regions had a political breakdown of 2e-9R. There were no districts that favored Dems at all. In the HD plan the same area was 1D-1d-3e-1r-16R. Not only are there 2 districts now that favor Dems, but one is solid D. In term of SKEW the net effect was to shift 3 points to the Dems.

In most of the regions of eastern WA there was no effect. Both parts of the Ellensburg-Moses Lake LD are just a solid R as the whole. Same thing with all four parts of the 2 LDs in the Tri-Cities area where the Hispanic population doesn't (or can't) vote and the rest is very Pub. The one lean r HD comes from the effect of Washington State Univ in Whitman county.

The shift in the distribution comes from the two large cities in the east: Spokane and Yakima. Spokane county has a population for about 3 1/2 LDs or 7 HDs. The city itself has more population than 1 LD, so it was divided along the Spokane river. The part of the city north of the river was one LD (reds) and the part south plus the west and south school districts in the county formed the other LD (cyans). Both of these were politically even with PVIs of 0 or 1. However when they were split the light red is a lean D (D+4) and the light cyan is a solid D district (D+9), leaving the dark red HD even (D+0.1) and the dark cyan HD as solid R. It is interesting to note that adding the reds together only creates a D+1 LD, but that's because the dark red provides 50% more votes than the other half and dominates the result.



The other significant change in skew comes from Yakima. Both the salmon and green combined LDs are solid R. However, the light salmon HD in the Yakima school district part of the city is an even political district, though it has half the voter turnout of the West Valley school district part of the city due to the high Hispanic population in the east part of the city. The green HD north in NE Yakima county has an even higher Hispanic population (HVAP 52%) than the Yakima city HD (HVAP 42%), but lacks the urban white Dem crossover vote of the city, so the green HD stays solid R. The light green HD in south Yakima county includes Klickitas and Skamania and is 31% HVAP, but also is 9% NVAP from the Yakama reservation. Those coalition minority votes are enough to make it an even political HD.

The light red HD in NE Spokane would contain Gonzaga University along with several of the poorest ZIP codes in Washington State - no wonder the turnout is anemic there.

That is, furthermore, one of the better splits of Spokane I've seen. The river and Division Street make natural dividing lines. I'm curious what the HD border within the cyan district is, it isn't clear from the map.

Ideally I would just follow Hangman creek, but that leaves the dark part about 5000 people short compared to the light part. To balance the population I shifted the Comstock area south of 29th and west of Perry. Technically the map shows the area east of Grand and north of 37th staying in the light area. That's because the precinct west of the creek includes the neighborhood between the creek and the tracks north of 14th so the dark area has that extra population. If precincts could be split I would probably have swapped those areas.