redistricting in Taipei City

Rather than continue to write what is proving not to be the most inspired professional paper of my career, I thought I’d take a break and write something for ye olde bloggue.  And while I could write about who is running and who is not, reporting on breaking news is not really my forte.  If you haven’t noticed by now, this blog is always at least a couple of days out of date.  That’s partially by design.  Anyway, enough blathering.  Onto redistricting in Taipei City!

(If you haven’t read the post on redistricting in Taipei County, you might want to do that first.)

Remember the rules.  Populations can’t deviate from the mean by more than 15%.  You can’t cut up administrative districts unless they exceed 115% of the average electoral district.  Everything has to be contiguous.  And you have to take into account geography, transportation networks, historical legacies and such.  In other words, you can’t stick strange areas together.

That second rule, about cutting up administrative districts, is useless in Taipei.  Taipei has 12 administrative districts, and 8 electoral districts.  None of the administrative districts is big enough to cut by that rule, but you can’t draw eight equally sized districts without cutting something.  So basically you are free to do as you please on that count.  This gives you a lot of freedom to maneuver.  When I discussed Taipei County, the size of townships eliminated most of the designers’ freedom to produce different sorts of plans.  In Taipei City, you could go in lots of different directions.

The plan produced by the Taipei City Election Commission sailed through the political process with minimal challenges, and this is the plan that we live under today.  This plan has clear advantages for the KMT.  (Note: I use KMT and blue camp interchangeably  Same for DPP and green camp.)

First, look at how reasonable the lines appear at first glance.  These are not weird shapes.  They basically follow administrative lines or cut districts into recognizable pieces.  If you are inclined to scream unfair, remember that before we start digging into the politics, this plan doesn’t hit you in the face with obvious political manipulation.

Second, a little about basic partisan geography.  The KMT is dominant in the southern part of the city.  It doesn’t matter how you arrange Wenshan, Da-an, Zhongzheng, and Xinyi; the KMT will have an easy majority no matter what you do.  The KMT also has a big majority in Neihu.  The DPP is stronger along the river (the western border) and toward the north.  The DPP only has a clear majority in Datong and in the Shezi part of Shilin.  However, there are lots of areas that can go either way.  Most of Wanhua, Zhongshan, the rest of Shilin, Beitou, and smaller parts of other areas are basically tossup areas or only lean slightly to the KMT.  So the real question is how to draw the lines in northwest half of the city.

Third, for illustrative purposes, I’m going to use Su Tseng-chang’s vote in the 2010 mayoral race.  In my academic stuff, I’m looking at the 2004 legislative race since that’s what the planners had at hand.  However, here I’m going to do things a bit differently just for fun.  It doesn’t really make much difference.  Partisan patterns in Taipei City are very stable.  Anyway, Su’s election is not quite a high water mark for the DPP, but it is a very good result.  If they are going to win seats, they aren’t going to have many more votes than this to work with.    I’m also going to be using different numbers to estimate population sizes.  The official plan lists the populations of each district, but I don’t have access to the population numbers at the li level, so I’m going to be using eligible voters.    (Yes, all very quick and dirty.)

Note how the official plan made really small districts where the KMT was strong:

Su % Eligible voters Official population
D1 47.6 261935 2.4 334363 2.7
D2 55.1 255988 0.0 325598 0.0
D3 44.0 276678 8.2 345086 6.0
D4 42.2 293352 14.7 371665 14.1
D5 46.6 239425 -6.4 307963 -5.4
D6 37.5 241754 -5.5 311626 -4.3
D7 41.1 240406 -6.0 308313 -5.3
D8 35.0 236387 -7.6 300300 -7.8

D6, D7, and D8 are all KMT strongholds and undersized, and they seem to have gotten even smaller over the past four years (or maybe there is something systematically different between population in 2006 and eligible voters in 2010).  But this isn’t too bad, and the biggest district (D4) is also a fairly safe KMT district.  So maybe I’m making too much of this except in one case, which I will discuss below.

You can see why all the DPP politicians want to represent D2.  It’s the only one that Su won, and he won it handily.  D1 and D5 are within shouting distance, but they still lean clearly to the KMT (and remember that this was a good year for the DPP).  They have to be considered unlikely, though possible, for DPP candidates.  The other five districts are probably impossible.  In the parlance of redistricting, this is a classic packing plan.  The KMT took the DPP’s very best areas and put them into one district, effectively sacrificing that district (their victory in D2 in 2008 notwithstanding).  However, all the other DPP areas were diluted to the point that the KMT should be able to win all seven of the remaining districts.

So let’s see what the DPP could have done to better its lot.  This should also illustrate how the KMT created its masterpiece.  There are some minor changes that would have been well within the spirit of the process and could have been adopted.  I’ll also put together my “evil genius” plan, a no-holds barred American-style plan full of unlikely combinations that would never have passed here.  It’s shocking to see just how much the DPP left on the table.

Minor Change #1: Shifting population between D3 and D7

Zhongshan District is usually thought of as one of the DPP’s best areas in Taipei City.  It combines with Datong to form a city council district, and the green camp routinely beats the KMT in votes and seats in this district.  One obvious DPP demand was for that city council district to simply form a new legislative electoral district.  After all, it is the right size.  However, doing that leaves an awkwardly sized population in Shilin and Beitou to the north.  It might be possible to make Zhongshan and Datong into one electoral district, but I haven’t figured out how to do that in any reasonable way.

Instead, the KMT plan combined Zhongshan, Songshan, and Xinyi districts to form two electoral districts.  Let’s stick with that, because a minor change that wouldn’t require changing any other districts could have important partisan impact.  Currently, the Songshan is split along Nanjing E. Rd, with the areas north going to D3 (Zhongshan) and the areas south going to D7.  That dividing line is very strategic.

Su Eligible
D3 Zhongshan 47.4 176610
Songhan A 38.1 100068
Total 44.0 276678
D7 Xinyi 39.9 179925
Songshan B 44.6 60481
total 41.1 240406

There are two things here.  Zhongshan is close, and Xinyi is not.  So the KMT put as much of Songshan as possible into D3 in order to dilute the DPP strength there.  Note that Zhongshan and Xinyi have roughly equal populations, but roughly 5/8 of Songshan went into D3.  More importantly, they chose the best KMT areas from Songshan to put into D3.  This meant strengthening the DPP a bit it D7, but that hardly matters.

Geographically, Songshan is like a box within a box.  Right in the middle of the district, there is a box (roughly the Minsheng Community area) with very strong KMT areas.  In the outer box, the partisan balance is more like Zhongshan, with only a slight lean to the KMT.  A strategic DPP planner could have drawn a vertical line roughly along Guangfu N. Rd. (to the south, Guangfu is the border between Da-an and Xinyi) that would have put most of that inner box into D7.  This would also have reversed the population ratios, with D3 now being about as small as the original D7.  This shift would increase Su’s vote share from 44.0% to 46.2% in the new D3.  This wouldn’t have created a “good” DPP district, but D3 would have become a significantly more possible district for the DPP.  A 4.4% swing is nothing to sneeze at.

Su Eligible
D3 Zhongshan 47.4 176610
Songhan A 43.2 65306
Total 46.2 241916
D7 Xinyi 39.9 179925
Songshan B 38.6 95243
total 39.5 275168

Note: If you want to see this on a map, go to Huang Chi’s wonderful site:

He has all the electoral data put on maps down to the li level.  I would should you myself, but I can’t figure out how to cut and paste stuff from there into this blog.

Minor Change #2: Shifting population between D1 and D2

The second change is very similar.  Currently, Beitou, Shilin, and Datong combine to form two electoral districts.  Overall, the two districts lean slightly to the DPP, but instead of two good DPP districts, there is a pro-KMT district.

There is a clear partisan gap between the areas from Shilin in D1 and D2, and this is clearly intended to make the KMT stronger in D1 and the DPP stronger in D2.  Now, this current division is certainly defensible.  The Tianmu area fits very well next to Beitou.  It conveniently also happens to be one of the KMT’s best neighborhoods in Shilin.  (The best would be just south of this one, between it and Zhongshan.  We will revisit this neighborhood in the super evil plan.)

Su Eligible
D1 Beitou 48.7 194177
Shilin A 44.6 67758
Total 47.6 261935
D2 Datong 57.6 98808
Shilin B 53.6 157180
total 55.1 255988

What if instead of the Tianmu area, we switched the Shezi Peninsula to D1.  Shezi is the DPP’s strongest area in the entire city, and this could tip the partisan balance.  Now, one can certainly argue that Shezi has more important transportation links to Datong than to Beitou, but let’s ignore that.  Here’s what we could get:

Su eligible
D1 Beitou 48.7 194177
Shilin A 62.5 51094
Total 51.6 245271
D2 Datong 57.6 98808
Shilin B 47.3 173844
total 51.1 272652

Sure the population is a little unbalanced, but it is easily within acceptable tolerances.  And presto!  There are now two tossup districts!  (Let’s note that there are almost certainly a few DPP politicians currently based in D2 who wouldn’t have been happy to make this trade.)

Minor Change #3: Shifting population between D5, D8, and D6

This is a change that was not made.  However, it was proposed late in the redistricting process, and it lost out in the lottery.  It was proposed by KMT legislator Lin Yu-fang, and it would have helped the KMT and him personally.  (So it doesn’t really belong in this essay, but I think it’s interesting.)

In the final plan, Wanhua, Zhongzheng, and Wenshan combine to form two districts (D5 and D8), while Da-an is its own district (D6).  All three are undersized.  Let’s look at it from Lin’s perspective in D5.  Wanhua is one of the DPP’s traditional strong areas.  It is gradually changing to a more neutral balance, but it remains a place where the DPP could put up good results.  Fortunately for the KMT, Wanhua is not big enough to form its own district, and the areas to the east and south are very strong KMT areas.  They added most of Zhongzheng to Wanhua to form a district that the KMT should almost always win.  However, this district is still undersized, and Lin would have preferred to add all of Zhongzheng.  This would not have violated any population requirements for D5, but it would have caused a problem for D8.  Lin’s proposal was to take the necessary areas from D6, which would have left both D6 and D8 right at the lower threshold for population requirements.  It wouldn’t have had much partisan impact on D6 or D8 since both are overwhelmingly pro-KMT, so we’ll ignore that (and save me the time of calculating those numbers).  However, this change would have solidified the KMT in D5.  Here is the actual plan:

Su eligible
D5 Wanhua 49.7 153932
Zhongzheng A 40.7 85493
Total 46.6 239425
D6 Da-an 37.5 241754
D8 Wenshan 34.3 200983
Zhongzheng B 39.2 35404
total 35.0 236387

Lin’s plan would have created a better D5 for the KMT.  Note that the part of Zhongzheng currently in D8 is actually the strongest KMT part of Zhongzheng.  Lin really would have liked that to be in his district!  Here’s what the new D5 would have looked like.

Su eligible
D5 Wanhua 49.7 153932
Zhongzheng 40.2 120897
Total 45.6 274829

D6 and D8 would have had an average of 221369 eligible voters, which would have been 13.4% below the mean.

This is a change that would have helped the KMT.  The equivalent change to help the DPP would have been to shift 20,000 more voters from D5 into D8, drawing the line in Zhongzheng a bit further north.  However, since the district is already undersized, you couldn’t do too much more.  Remember, the voters north of the line are already closer to the overall average than the ones to the south, moving half as many would not have that much impact.

Evil Change #4: Cut Youth Park area out of D5

Now we start to get into some ideas that are pushing the envelope for what might be considered acceptable and what moves into naked partisanship.  As we noted above, the DPP is fairly strong in Wanhua.  However, this is not true in the neighborhood around Youth Park, which has a lot of public (read: military (read: mainlander)) housing.  Conveniently, this borders the area of Zhongzheng that is already in D8.  What if we cut out those areas and replaced them with some votes from Da-an?  We’ll take the blocks between Xinsheng S. Rd. and Fuxing S. Rd. and between railroad and Xinyi Rd; this basically extends the district east two blocks.  We won’t worry about how to replace those areas in D6.

Here’s the current D5:

Su eligible
D5 Wanhua 49.7 153932
Zhongzheng A 40.7 85493
Total 46.6 239425

Here is what the new district would look like:

Su eligible
D5 Wanhua 54.4 124471
Zhongzheng A 40.7 85493
Da-an A 39.1 29095
Total 47.8 239059
To D8 Wanhua 30.3 29461

These are all minor shifts, but when you are so close to the 50% mark, these minor shifts matter a lot.

Super Evil plan!

Ok, this is the no-holds barred, I have no sense of shame, naked partisan advantage plan.  This is what I would come up with if I were an American planner (where all those adjectives are routine and expected) trying to maximize DPP chances at success.  Here are my eight districts:

Su eligible
D1 Beitou: all 

Shilin: Shezi Peninsula

51.6 245271
D2 Shilin: most of Shilin 

Datong: north of Minquan W. Rd.

Zhongshan: north of Minquan E. Rd. but not including Dazhi area

51.7 242116
D3 Wanhua: not including Youth Park area 

Datong: south of Minquan W. Rd

Zhongshan: south of Minquan W. Rd, but not including area south of Minsheng E Rd. and east of Songjiang Rd.

53.7 240729
D4 Shilin: a few areas near National Palace Museum and up toward Tianmu (but not Tianmu itself) 

Zhongshan: Dazhi

Neihu: most of Neihu

Songshan: Minsheng Community east to river

38.2 270510
D5 Neihu: Donghu and area along Jilong River 

Nangang: all

Xinyi: areas along Nangang and Sonshan borders

Songshan: areas near Ciyou Temple

45.3 238368
D6 Songshan: north and west sides 

Zhongshan: area south of Minsheng E Rd. and east of Songjiang Rd.

Zhongzheng: old Chengzhong District

Da-an: north of Xinyi Rd.

41.2 269213
D7 Da-an: south of Xinyi Rd. 

Xinyi: most of district

36.6 273870
D8 Wanhua: area around Youth Park 

Zhongzheng: most of old Guting District (same as current plan)

Wenshan: all

34.5 265848


D1 is the district I proposed above.  So far nothing too radical.  D2 and D3 are beyond the pale.  Basically, we have to put together Datong, Zhongshan, Wanhua, and the rest of Shilin to form two districts.  There simply isn’t any “reasonable” line that I can find that does this.  And there are geographical problems.  To the south, Wanhua only touches Datong.  We really don’t want to wander into Zhongzheng to get to Zhongshan, because the DPP’s vote is 10-15% lower on the other side of the street.  Shilin really also has to connect through Datong.  Shilin and Zhongshan are connected, but I cut out those places from both districts and put them into D4.  So my solution is to draw a horizontal line right through the middle of Datong and Zhongshan, putting part of each administrative district in D2 to the north, and half in D3 to the south.  So both D2 and D3 contain areas from three administrative districts but no complete administrative districts.  Moreover, I cut out the DPP’s bad areas in each of these districts.  (As an American, I can draw these districts without so much as blushing!)

D4 gets all the non-competitive areas in the northern part of the city.   This district includes most of Neihu, except for the very southern and eastern sections, and then it stretches west through Dazhi (the part of Zhongshan north of the river) and through the tunnel into Shilin, near the National Palace Museum and a bit north.  I also really pushed the envelope by crossing the river on the eastern boundary of Songshan District in order to take in the Minsheng community area (but only that area and not the greener areas in Songshan to the north or south of it).

D5 is my attempt to put together all the remaining areas in Taipei City that the DPP doesn’t get steamrolled in.  This includes all of Nangang.  In Neihu, it includes the of Neihu along the river (on the southern border) and on the eastern border, next to Xizhi.  These areas are generally not very glamorous.  The area along the river is some of the most industrial grit you can find in our fair city.  In Songshan, the area right around the Ciyou Temple and Raohe Night Market is pretty good DPP territory.  This is due south of the part I put into D4.  Finally, the parts of Xinyi district on the northern and eastern borders are ok for the DPP.   Roughly, you can think of this as areas near the Houshanpi MRT station.

D6, D7, and D8 are hopeless for the DPP.  I tried to put the best of the remaining areas into D6, but there isn’t much left.  You have the northern and western parts of Songshan, the southeast corner of Zhongshan, most of Zhongzheng (the part of Zhongzheng  that is in the actual D5), and the northern part of Da-an (just extend Zhongzheng eastward).  D7  is the rest of Da-an and all the parts of Xinyi that we didn’t want in D5.  D8 is the same as the actual D8, except we added the areas from Wanhua right around Youth Park.

The population distribution is fine.  The biggest district is 7% above the mean, and the smallest is 7% below it.  I could have gone further here.  Regardless, you will notice that the D1, D2, D3, and D5 are smaller than D4, D6, D7, and D8.

What kind of monster have I created!?!?

So here you would have one district with the DPP as a clear favorite (D3), two tossups or perhaps DPP slight advantages (D1, D2), and a fourth district that the DPP has an outside chance in if absolutely everything goes right (D5).  The other fourth districts are impossible for the green camp.

Of course, to get this you have to violate every sense of following the spirit of the redistricting guidelines.  Oh, I could have been even more egregious, I suppose.  I could have taken the one li in Nangang that borders Wenshan district and changed it from D5 to D8 even though there is only one little mountain road (that goes through a graveyard and a garbage incinerator) connecting them.  But I didn’t leave too many opportunities like that left untouched.

How do I feel about this?  In some ways, I’m stunned that the DPP allowed themselves to get rolled like this.  Even little changes, like my proposed changes #1, #2, and #4 would have created a significantly better electoral environment for them.  #4 might be crossing the line of what is acceptable, but #1 and #2 certainly are not.  It would not have been difficult to propose an alternate plan to the Speaker (Su Tseng-chang).  At the very least, they would have had a 50-50 chance in a lottery of getting a better set of districts.

Now, we could just assume that the DPP was stupid and naïve and got beaten on by the much more sophisticated KMT.  I don’t like to make these kinds of assumptions.  To me, any time you find yourself explaining something with the rationale that “people are stupid,” you probably need to rethink.  That is usually an excuse for lazy thinking on your part.  So I’m trying to find reasons that the DPP would have been against changes like the ones I outlined.  With #4, there might have been sufficient commitment to ideas of playing fair, that spanning three districts and obviously cutting out a bad area just might not have been acceptable.  Besides, any politicians wishing to run in D6 probably would not have been happy with such a switch.  With #2, there were certainly DPP politicians (ie: most in Datong and Shilin) who would have been unhappy with the switch.  #1 is the one that I think is the hardest to explain.  The only thing I can speculate about is that someone in Songshan wants to run, and they would like to be in the Zhongshan District.  However, I don’t know who this would be.

My other reaction is that I’m relieved that they didn’t allow someone like me to draw the districts.  It’s better for democracy if you don’t look at a district and immediately assume that it was drawn in this funny way for some political purpose.  However, this is always a possibility in a system like Taiwan has now, with single-member districts.  As long as you have to draw districts, the possibility of partisan manipulation is going to be present.  To me, this is a very strong argument for electoral system change.  I have other arguments as well.  I generally think that this electoral system is a disaster on many fronts.  However, even if this were the only problem, it would be sufficient to consider a different system.

3 Responses to “redistricting in Taipei City”

  1. Ethan Wu Says:

    Here are some maps for helping illustration. Just make for fun!!

  2. frozengarlic Says:

    Wow! How did you do that? Are you using GIS software? That is very close to what I envisioned, which is no small feat considering how vague my description is. Very cool!

  3. Ethan Wu Says:

    Actually I was calculating something by my ArcGIS about the election this year, and then I found this interesting article. Though I’m a Taiwanese (actually Taipeinese), lots of information are still out of my knowledge. Then I though why not just easily make a map, not only help myself but more followers of this blog to understand what you are talking about.

    I use “li” as the basic unit. And the boundaries are following two rules.
    1. If the article mentioned a road specifically, then I just followed the road as much as possible.
    2. If the article just mentioned an area roughly, then I followed the boundary of “sub-division” (次分區), which is developed by city government.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: