I’ve been working a lot on the redistricting process from 2005-7, and one of the most striking things about the process was just how non-partisan it was. This is not to say that it was completely non-partisan. In fact, you find numerous indications of struggles for partisan advantage. What I’m trying to say is that I expected a lot more. Instead, almost every district (and indeed almost every proposal that was not eventually adopted) looks like it could have been drawn up by some person with detailed knowledge of local history, geography, economics, and transportation networks but absolutely no idea of the partisan balance. In short, the designers showed remarkable restraint, possibly because they were electoral commissions and had to explain all their decisions without reference to partisan strength.
One of the mental exercises I’ve been doing is to throw all that restraint out the window and try to design a more advantageous plan for one party or another. Sometimes this is really easy (eg: producing a better plan for the DPP in Taipei City), and sometimes it is extremely hard. Here, I want to look at one of the hard cases.
Taipei County had 12 seats, the most in the country. You would thus expect that there are lots of different variants, some better for the KMT and some better for the DPP, to choose from. My theory says that, due to local conditions, the plan chosen should have benefitted the KMT. This means that I should be able to see, or at least imagine, an alternate plan that is better for the DPP. So let’s try to produce that plan.
First, there are some rules. You have to respect geography, transportation networks, historical influences, and so on. (This is as vague as it sounds.) The population of any district can’t exceed 15% from the county mean. Taipei County had 3698674 people, which means that each district should have an average of 308223 people. So each district has to be between 85% (261990) and 115% (354456) of that mean. You also have to respect township boundaries. Unless a township exceeds 354456, it should not be divided between two electoral districts. This restriction is not absolute, but you should try very hard not to violate it. Four townships (Banqiao 板橋, Xinzhuang 新莊, Sanchong 三重, and Zhonghe 中和) have to be split. It goes without saying (because it was so self-evident that they didn’t actually say it in the law) that districts must be geographically contiguous. There is also one more rule that no one ever mentioned that I think was implicit: no township should be split into more than two pieces.
So here is the final plan.
It turns out that it is extremely hard to come up with a fundamentally different plan. Yes, you can jigger a few lines here and there and switch a few voters, but it is very hard to design anything that looks completely different from this plan.
To understand why this is, start in Yonghe 永和 (in the 7 o’clock position). Yonghe has a population of 234381, so it needs a few more people to get to the minimum population. Since Yonghe only borders one township, Zhonghe, it is pretty obvious where these people will come from. Conveniently, Zhonghe is too big to be its own district and needs to be split. So the plan combines 69523 people in Zhonghe with all the Yonghe residents to form District 9 (D9), and all the other 337592 people in Zhonghe form D8. Now, you could take a different 70000 people from Zhonghe, but both districts are so heavily pro-KMT that it wouldn’t matter. (In fact, they appended the bluest areas of Zhonghe to Yonghe, which is even bluer. So this plan is relatively good for the DPP, except that the DPP has very little chance of winning even this Zhonghe district.) So, given our rules, you basically can’t avoid these two districts.
However, it turns out that this start goes a long way to determining the rest of the districts because Zhonghe links the townships to the east and west. Now, on the map, Xindian 新店 (to the east) borders both Tucheng 土城 and Sanxia 三峽 (to the west), but those borders are all mountainous. On the ground, everything goes through Zhonghe. In other words, you can’t combine Xindian with Tucheng or Sanxia.
So let’s move east, going counterclockwise around Taipei City. There are three huge puzzle pieces: Xindian (284997), Xizhi 汐止 (172909), and Danshui 淡水 (127140). Everything else to the north, east, and south of Taipei City is relatively small. None of the big three townships can go into the same electoral district, and none of them are big enough to split into two districts. So you have to have one electoral district centered around each of the three townships. So this means that you split the small townships in the SE region somewhere between Xindian and Xizhi to form D11 and D12. Politically, it really doesn’t matter where you do this, since Xindian and Xizhi are so heavily blue that they simply overwhelm the green tendencies in the smaller townships.
The only flexibility is on the eastern border of the Danshui district (D1). (Note: Danshui is the township on the north coast on the east bank of the mouth of the river.) There are four townships between Danshui and Xizhi. You can include as many of these as you like in D1. So, to the east side of the river, D1 can have between 127140 and 203725 people. Either way, we have to add some more. Let’s go ahead and add 31527 from Bali, across the river. Now we have between 158667 and 235252. Still not enough.
Let’s stop for an aside from a Xinzhuang-centric 新莊 viewpoint. I do this both because a politician based in that area screamed bloody-murder about the unfairness of the plan, and also because I think that this is probably how most people look at redistricting. If you don’t think about the rest of the county, then it is quite obvious how things should be divided from the Xinzhuang perspective. Xinzhuang is a big city on the north bank of the Da-han River 大漢溪, and it serves as the major economic center for the townships to the north. If you combine the populations of Xinzhuang, Wugu 五股, Taishan 泰山, and Linkou 林口, you get 586888, which is enough for two districts. If you add Bali 八里 (though Bali residents might argue that they have closer ties to Danshui), you get 618415, which is almost perfect. So obviously, there should be one district with Wugu, Taishan, Linkou, Bali, and part of Xinzhuang, and another district made up from the rest of Xinzhuang. Wonderful! This even leaves a viable pair of districts for Sanchong 三重 and Luzhou 蘆洲, which together have 566882 people. Unfortunately, as we have already discovered, this won’t work because of the areas on the north and east of Taipei City. You would have to split Xizhi, and Xizhi isn’t big enough to be split. So for mechanical reasons, what seems perfectly reasonable from the Xinzhuang-centric point of view turns out not to be viable.
Let’s go back to D1. D1 has Danshui, Bali, and zero to four of the smaller townships on the north coast. To meet the population requirement, we must add one or two of Wugu, Linkou, and Taishan. Hold that thought.
We started in Zhonghe and moved counterclockwise. Let’s go back and move in the other direction. Banqiao 板橋 is to the northwest of Zhonghe. Banqiao has 539534 people. You can legally split this into two districts and be within the 15% limit, but they are pretty small. We’d like to add some people if possible. Unfortunately, it’s not clear where these people should come from. To the north of Banqiao is the Da-han River. This is a big geographic barrier, and it seems to violate common sense to add areas from the other side. To the west and south are Shulin 樹林 and Tucheng. Neither is big enough to split. To the southeast is Zhonghe, which could spare 30000-40000 people. However, Zhonghe has already been split, and this would require splitting it into three pieces. Imagine the outrage. All of these options have obvious flaws, and the obvious thing to do would simply be to split Banqiao into two very small districts. That is what eventually happened to form D6 and D7.
Wait, there’s something else. There is a partisan dimension to having small Banqiao districts. The areas that might be added could affect the balance of power in the new D6 and D7. The DPP is a lot stronger in Xinzhuang and Sanchong across the river, but adding people from those areas is highly unlikely. More interestingly, the areas of Tucheng and Zhonghe that abut Banqiao are the DPP’s strongest areas in those cities. The Zhonghe areas are slightly more pro-KMT than Banqiao and the Tucheng areas have a slightly more DPP tilt. This matters because Banqiao is currently almost a tossup. Moving the balance just a bit one way or the other could be decisive. Now, the Tucheng areas come out of another tossup district, so the only question there is whether you would rather win one district or another. However, putting the Zhonghe areas with Banqiao would strengthen the KMT in both Banqiao and D8 (which would be losing the only areas in which the two parties are even close and thus become even more overwhelmingly blue). However, it’s a small shift, and it would require splitting Zhonghe into three pieces.
So if you aren’t going to do any of that, and you just keep the two small Banqiao districts, the next step is obvious. Tucheng and Sanxia almost have to form D10, since there is nothing left unused to their southeast. Fortunately, they jointly have 319876 people, which is almost ideal. However, the next townships over, Yingge 鶯歌 and Shulin, don’t have quite enough people (238112) to form their own district (D5). This means D5 needs a few more people, and Xinzhuang is the only available source. The rest of Xinzhuang then forms D4.
This brings us back to D1, D2, and D3. Sanchong is too big and has to be split. So most of Sanchong will form D3, and the rest can only go across the river to Banqiao (which we have already dismissed as unrealistic), west to Xinzhuang (which already has enough people), or north to Luzhou. Conveniently, Luzhou and Sanchong have close ties historically, economically, transportationally, politically, etc-ly. Sanchong and Luzhou have enough population to form two undersized districts, but D1 can only include two of Wugu, Taishan, and Linkou. Since Wugu adjoins Luzhou, it has to go into D2 while the other two go into D1.
Shockingly, there is very little room for maneuver in all this. Nearly every significant change entails breaking one of our explicit or implicit rules. You can fiddle around with some of the minor details, but it is very hard to come up with something fundamentally different.
(For reference: In the final stages of the redistricting process, there were in fact two plans, one supported by the green camp and one supported by the blue camp. However, these two plans were almost exactly identical. The only difference was that the green camp’s plan shifted 9 li (42351 people) from Xinzhuang into D5 and the blue camp’s favored plan shifted 13 li (61957 people). This had almost zero partisan impact. They decided by drawing lots. The former plan won.)
There are really only two fundamentally different alternate plans that I can see. Late in the process, I saw a newspaper report of a KMT legislator who suggested dividing Tucheng. I never saw another mention of this plan, so it must have been squashed. However, you can see how this might have worked:
D6: 316000 (the rest of Banqiao)
D7: 316074 (Banqiao: 223534, Tucheng: 92540)
D10: 309741 (Tucheng: 140000, Yingge, Sanxia)
D5: 304770 (Xinzhuang 80000, Shulin, Taishan)
D4: 305480 (the rest of Xinzhuang)
D1: 293197 (all townships on north coast, Linkou to Wanli)
D12: 282756 (Xizhi plus all townships to SE of Taipei)
D11: 287878 (Xindian and Wulai)
D8, D9, D2, D3: no change
What kind of partisan impact would this have? D6 and/or D7 would probably be slightly more prone to go to the DPP than in the actual plan. D10 would probably be about the same. It would lose some mildly pro-DPP areas to D6 and D7, but it would gain Yingge, which is also mildly pro-DPP. D5 would effectively be trading Yingge for Taishan and a little more of the western tip of Xinzhuang. This is probably a slight gain for the KMT. D1 would lose Taishan but gain Jinshan and Wanli. That’s probably a wash. D11 and D12 are too blue to worry about any minor changes. Overall, the partisan impact would be mild, though it looks to me like it would probably benefit the DPP more than the KMT. The bigger impact would be on which politicians from each party were advantaged. For example, in the current D10, Tucheng makes up about three-fourths of the district. That makes it all but inevitable that the representative will come from Tucheng. In the alternate plan, Tucheng is only about half of the district, and a good chunk of Tucheng is buried inside one of the Banqiao districts. This makes it a lot harder for an ambitious Tucheng politician to win a legislative seat.
The other alternate plan is much simpler. Currently Banqiao City is divided into a northern and southern district. Earlier, I said that they were both tossup districts, but that isn’t quite right. After the Tsai Ing-wen’s strong performance in the 2010 election, they look like tossup districts, but from a longer term perspective, the KMT should be slightly favored in both. Even more precisely, the KMT has a very miniscule edge in the northern district (D6) and a somewhat larger edge in D7 (where the KMT is roughly 3% stronger). Tsai Ing-wen got 50.3% in Banqiao, so I’m guessing she won about 49% in D7 and 52% in D6.
Now, since Banqiao has to be divided, it can be divided in pretty much any way you like. If I wanted to put as many DPP votes as possible into one district and as many KMT votes as possible into the other, I would slightly tilt that axis and have a northeast and a southwest district. By doing that, you could shift about each district by about 2%. That is, instead of being 3% different, the two districts would be more like 7% different. Instead of having two districts that are roughly similar, one would be clearly better for each party. To put it another way, looking from the lens of 2006, in the actual plan the KMT seems to have a slight advantage in both districts. If we redivided Banqiao, the KMT would be favored to win one district easily, but the DPP would have a good chance of winning the other district. If I were a local KMT politician in 2006, I would want two good districts to choose from. Likewise, if I were a local DPP politician, I would want one solid district rather than two possible but not likely districts. However, if I were a national party strategist, I might want the opposite. If the DPP wants to win a majority in the legislature, it probably has to win both of these districts. By making them both roughly the same, the planners marginally increased the DPP’s chances of taking power in the legislature.
(In the earliest stages of the redistricting process, the Taipei County Electoral Commission suggested a plan that divided the city along Zhongshan Rd 中山路 instead of Xianmin Blvd 縣民大道 (ie: the railroad). This would have been roughly halfway between the eventual plan and my completely politicized Frankenstein plan. However, that plan was dropped.)
With 12 districts, this doesn’t seem like a whole lot of scheming. Of the three viable plans that I can think of, the eventual one was perhaps the most advantageous for the KMT. However, these advantages are surprisingly slight. Given the original rules, there really isn’t much to complain about for either side.