by-election results

The results of the by-elections are in.  On turf that heavily favors the KMT, the DPP won three of the four seats.  Here’s the breakdown:

English name Chinese name Party Votes Percent
Taoyuan County 3
Huang Renzhu 黃仁杼 DPP 45363 47.3
Chen Xuesheng 陳學聖 KMT 42600 44.4
Wu Yudong 吳餘棟 IND 4424 4.6
Lin Xiangmei 林香美 IND 3617 3.8
Hsinchu County
Peng Shaojin 彭紹瑾 DPP 71625 56.0
Zheng Yongtang 鄭永堂 KMT 56342 44.0
Chiayi County 2
Chen Mingwen 陳明文 DPP 57451 67.9
Lin Derui 林德瑞 KMT 27138 32.1
Hualian County
Wang Tingsheng 王廷升 KMT 39379 48.3
Bi-khim Hsiao 蕭美琴 DPP 33249 40.8
Shi Shenglang 施勝郎 IND 8863 10.9

For the record, my predictions were:

Chiayi County 2: Chen Mingwen 陳明文 (DPP) 69; Lin Derui 林德瑞 (KMT) 31.

Hsinchu County: Peng Shaojin 彭紹瑾 (DPP) 54; Zheng Yongtang 鄭永堂 (KMT) 46.

Taoyuan County Third District: Chen Xuesheng 陳學聖 (KMT) 49; Huang Renzhu 黃仁杼 (DPP) 40; Wu Yudong 吳餘棟 (IND) 7, Lin Xiangmei 林香美 (IND) 4.

Hualian County:  Bi-khim Hsiao 蕭美琴 (DPP) 44; Wang Tingsheng 王廷生 (KMT) 42; Shi Shenglang 施勝郎 (IND) 14.

So I got Chiayi County 2 and Hsinchu County almost exactly right, and I messed up Hualian County and Taoyuan County 3.  Oh well.  I wasn’t too far off, and election predictions are hardly a scientific exercise anyway.

In retrospect, I should have paid more attention to myself when I was discussing the underlying partisan structures of Taoyuan 3 and Hualian.  Both seem to have changed from 2008 by similar amounts; Hualian was simply a bit more overwhelmingly blue to start with.

By the way, one of the hardest things to predict is what will happen to third-party candidates.  Third-party candidates often have wide gaps between their performance in the surveys and at the ballot box.  Some candidates’ support simply disappears, most likely due to strategic voting.  Others’ holds steady.  Once in a while, a candidate with low support in surveys will produce an eye-popping vote count.  From afar, it’s very hard to tell which will be the case with any particular candidate.  This time, I came very close in predicting the three independent candidates.  However, this was mostly luck.

So that’s what happened.  More importantly, what does it mean?  From watching the talking heads on TV last night, it is obvious that you can draw nearly any lesson you want to.  Some of the stories various people were pitching include 1) we are seeing a structural change in partisan attitudes, 2) negative campaigning doesn’t work, 3) the KMT reform isn’t working, 4) the KMT reform needs more time, 5) the KMT simply failed to mobilize its votes (but no real partisan realignment), 6) the DPP will sweep all five mayoral seats later this year, 7) Ma should resign as KMT party chair, 8) the KMT nominated poorly.  Do I buy any of these?  The short answer is 1) kinda, 2) no, 3) no opinion, 4) no opinion, 5) no way, 6) that’s a little premature, and 7) that’s silly.  The ones I want to try to dig into are the stories about partisan realignment and mobilization.  I think we are seeing a partisan shift, or perhaps a reversion to the alignment of 2003 or so.  I think the mobilization story is fundamentally flawed.

Let’s do mobilization first.  The clearest version of the mobilization story I have heard is this.  Last December, Peng Shaojin 彭紹瑾 ran for Hsinchu County executive and got 77,000 votes.  He lost because the KMT candidate mobilized 97,000 votes.  In this by-election, Peng Shaojin ran again and got 71000 votes.  In other words, Peng turned out nearly every person who voted for him last time.  (In fact, the DPP has gotten between 70000 and 80000 votes in Hsinchu County in numerous elections over the past two decades.)  However, the KMT candidate this time only got 56,000, only about half as many as in December.  In short, the DPP mobilized all its votes, while the KMT did not.  This story is then fleshed out with a narrative of a divided KMT with key elites declining to work for the candidate in this election.  The Hsinchu County KMT has two dominant figures, Qiu Jingchun 邱鏡淳 and Zheng Yongjin 鄭永金.  In December, Zheng refused to support Qiu as county executive, backing a third-party candidate instead.  Qiu won, but it wasn’t easy.  The KMT tried to paper over this division by nominating Zheng Yongjin’s younger brother, Zheng Yongtang 鄭永堂, for legislator.  Publicly, Qiu expressed his support, but the actual support never materialized.  As a result, Qiu’s supporters did not turn out to vote for Zheng, and the KMT lost.

The critical feature of the mobilization story is that there are no partisan changes.  DPP supporters still support the DPP, and KMT supporters still support the KMT.  The only question is whether the parties can persuade their supporters to show up at the ballot box.  Thus, the reason the KMT has done so poorly in these by-elections is that the turnout has only been around 40%, as compared to around 60% in the 2008 legislative elections and 2009 county executive elections and 75% in the 2008 presidential election.  Since no one thinks that turnout in any of the coming general elections will be 40%, we should only worry about what a 60-75% turnout election looks like: still very blue.  These by-elections are simply an aberration.

One way to attack this theory is by looking at survey results.  TVBS published a survey of Hsinchu County voters on Jan 26, 2010.  In the report, they produced the following table:

LY race
total Peng Shaojin (DPP) Zheng Yongtang (KMT)
total N=827 44 34
County Executive Qiu Jingchun (KMT) 38 32 51
Peng Shaojin (DPP) 15 89 4
Zhang Biqin 張碧琴(IND, supported by Zheng Yongjin) 17 37 44

So the 15% of respondents who reported voting for Peng Shaojin in the county executive race continued to express support for him by an overwhelming 89% to 4% margin.  So far so good for the mobilization story.  However, 32% of the 38% who said they voted for Qiu in December expressed support for Peng, and 37% of the 17% who voted for Zheng Yongjin’s candidate in December said they would vote against Zheng Yongjin’s brother in February.  This certainly sounds like Peng was picking up support from the KMT, not simply mobilizing his previous supporters.  In other words, Peng may have gotten 70,000 votes again, but they weren’t the same 70,000 voters.  A lot of them crossed party lines.

Another way to attack the mobilization thesis is to look at election returns.  In 2008, the KMT was enthused while the DPP was depressed, so the KMT should have had a marked advantage in mobilization over the DPP.  In 2009-10, the roles were reversed.

County 2008 LY list vote 2009 county executive
Blue Green Blue Green
宜蘭縣 Ilan 87613 83313 112469 133394
桃園縣 Taoyuan 471687 265842 396237 346678
新竹縣 Hsinchu 136854 45009 97151 77126
苗栗縣 Miaoli 157868 64985 181256 95469
彰化縣 Changhua 308182 225526 348341 276897
南投縣 Nantou 125074 72421 136951 107023
雲林縣 Yunlin 144559 142869 121832 229958
嘉義縣 Chiayi 104909 121225 128973 177333
屏東縣 Pingdong 183157 187474 185384 270402
臺東縣 Taidong 54553 20072 56354 50802
花蓮縣 Hualian 87332 27244 38603
澎湖縣 Penghu 19403 14754 22664 22069
基隆市 Jilong 95939 45290 86001 65673
新竹市 Hsinchu City 97260 57252 92667 68822
嘉義市 Chiayi City 54914 52659 69962 61268
金門縣 Jinmen 23555 1190 14269
連江縣 Lianjiang 3719 385 5404

There are three basic patterns.  In some counties, the blue camp’s vote fell while the green camp’s vote increased.  In Taoyuan, for example the blue camp produced 70000 fewer votes in 2009 while the green camp increased its total by 80000.  Hsinchu County, Yunlin, Jilong, and Hsinchu City also saw blue decreases and green increases.  Second, many places saw both camps increase their number of votes, indicating that both were better at mobilizing in 2009 than in 2008.  Ilan, Miaoli, Changhua, Nantou, Chiayi County, and Chiayi City followed this pattern.  Then there is the third pattern, in which the KMT vote did not change markedly from 2008 to 2009 but the DPP vote increased dramatically.  This was the case in Pingdong and Taidong.

How can all these patterns occur with a simple story about mobilization?  One could explain this by delving into the differences in the various county executive races.  For example, in Yunlin, a popular DPP incumbent was expected to win re-election easily, so the energized DPP voters turned out while the demoralized KMT voters stayed home.  Thus, the KMT vote fell while the DPP vote skyrocketed.  Unfortunately, the story was almost exactly the same in Pingdong, except the KMT vote stayed the same.  Moreover, Chiayi County also had a race that the DPP was expected to win easily, and the KMT vote was much higher than in 2008.   Thus, we are expected to believe that demoralized KMT voters stayed home in Yunlin, voted at normal rates in Pingdong, and voted at very high rates in Chiayi County.

The stories about counties the KMT expected to win easily are similarly contradictory.  In Miaoli, Changhua, Jilong, and Hsinchu City, the KMT was widely expected to win by comfortable margins.  In all of them, the KMT experienced a slight increase in its vote.  According to the mobilization story, we would conclude that the enthusiasm over good prospects in the county race outweighed the national factors making it harder for the KMT to mobilize.  Ok, but what about Taoyuan?  The KMT’s vote dropped by 75,000.  Maybe Wu Zhiyang 吳志揚 was so sure of victory that he declined to spend the money needed for a full mobilization.  If so, he seems to have been the only one to dare to attempt the non-mobilization strategy.  If non-mobilization were a viable electoral strategy, one would have thought that the winners in Miaoli, Yunlin, and Pingdong might have attempted it as well, but we see no evidence of this.  Moreover, in light of the two-by elections in Taoyuan, it is equally plausible that Wu fully mobilized but that the KMT has simply become less popular.

In sum, I simply don’t think the pure mobilization / no partisan change story is a viable explanation for what has happened over the past two years.

I’m not arguing that mobilization doesn’t matter or that it isn’t part of the story.  We all know that mobilization is critically important.  However, political scientists have done a terrible job at explaining exactly how mobilization matters precisely because it is very complicated.  Simplistic stories like the one I argued against are nearly nuanced enough.  They are very appealing though, precisely because they are simplistic.  They focus on one variable, the network of cadres that moves people to the polls.  If you put in more resources, you get more votes.  Taken to the logical extreme, this is an anti-democratic argument.  No matter what happens, KMT voters will continue to support the KMT.  It simply might be harder to get them to the polls.  However, the democratically produced president/legislator is a function of resources, not of appeals and/or actual public opinion.  As such, the legitimacy of such officials to act on behalf of the public should be discounted.  This story is only partially wrong, but that part is critical.  Mobilization and resources matter, of course, and can affect the outcome of an election.  However, voters do change their minds, so party platforms, issue appeals, and scandals also matter.  As a result, elections do not depend solely on resources.

The difficult part is figuring out how mobilization matters in conjunction with changes in opinion.  I accept the argument that it was harder to mobilize DPP voters in 2008 and harder to mobilize KMT voters in 2009-10.  I do think that a lower turnout rate helped the DPP this time.  However, I do not think that all DPP voters turned out and only part of the KMT voters did.  A higher turnout rate likely would have helped the KMT, but only slightly.  For example, the turnout in Chiayi County 2 was 38%, and Chen Mingwen 陳明文 won 68-32%.  If the turnout had doubled, Chen would almost certainly still have won handily, though not by quite as overwhelming a margin.  Perhaps it would have been something like 64-36%.  In that case, the half of the electorate that didn’t turn out (but would turn out in a presidential election) would have preferred Chen by a 60-40% margin.  Of course, that is just a guess.

We do have some evidence of the effect of different levels of turnout.  For example, the legislative election in Jan 2008 had a turnout of 58%, while the presidential election of March 2008 had a 76% turnout.  There was very little difference in the vote share of the camps.  The presidential election was 58.5-41.5%, while the legislative election was 55.2-40.4%.  Since independent votes tend to come mostly out of the blue camp’s vote share, the two are roughly equivalent.

Since we think that the KMT was enthused in 2008, their voters should have turned out more heavily in the first 60%, and the DPP should have benefited from the higher turnout.  Thus we have to invent some story about how Ma Yingjeou 馬英九 is a great candidate (or voters were sick of President Chen 陳水扁), candidate effects offset the turnout effect.  I don’t believe this.  I think the same partisan effects drove both elections.

After years of staring at voting returns, I have come to the tentative conclusion that turnout above 60% really doesn’t affect partisan results.[1] Another way of saying this is that almost everyone who can be mobilized has already been mobilized by 60%.  Higher turnout comes from voters outside the mobilization networks, and they tend to split their votes among the parties in nearly the same way that the first 60% do.  However, turnouts below 60% start to affect partisan outcomes.[2] In jargon, there is a non-linear relationship.

So, given that I think that there has been some partisan shift since 2008 and that turnout below 60% matters, where does the partisan balance stand right now?  Nearly half the electorate has voted in the past six months, so let’s compare them to a benchmark, the party list vote in the 2008 legislative election.[3]

2008 LY Party List 2009-10 LY by-elections
votes blue% green% valid KMT% DPP%
雲林縣二 Yunlin County 2 156911 47.2 46.6 126297 23.2 58.8
南投縣一 Nantou County 1 99695 62.8 31.9 119284 55.3 44.7
台東縣 Taidong County 78880 69.2 25.4 46887 45.2 49.5
台中縣三 Taichung County 3 138134 54.5 40.8 115111 45.0 55.0
桃園縣二 Taoyuan County 2 128994 55.3 39.2 92382 40.0 58.1
嘉義縣二 Chiayi County 2 112866 43.1 52.0 84589 32.1 67.9
桃園縣三 Taoyuan County 3 135277 63.7 31.8 96004 44.4 47.3
新竹縣 Hsinchu County 195144 70.1 23.1 127967 44.0 56.0
花蓮縣 Hualian County 119669 73.0 22.8 81491 48.3 40.8
total 1165570 59.8 34.8 890012 41.6 53.4
2008 LY Party List 2009 County executive
votes blue% green% valid KMT% DPP%
宜蘭縣 Ilan County 178872 49.0 46.6 245863 45.7 54.3
桃園縣 Taoyuan County 771850 61.1 34.4 758722 52.2 45.7
新竹縣 Hsinchu County 195144 70.1 23.1 252424 38.5 30.6
苗栗縣 Miaoli County 236260 66.8 27.5 284138 63.8 33.6
彰化縣 Changhua County 564959 54.5 39.9 634640 54.9 43.6
南投縣 Nantou County 208186 60.1 34.8 269217 50.9 39.8
雲林縣 Yunlin County 305439 47.3 46.8 351790 34.6 65.4
嘉義縣 Chiayi County 238892 43.9 50.7 317135 40.7 55.9
屏東縣 Pingdong County 386762 47.4 48.5 455786 40.7 59.3
臺東縣 Taidong County 78880 69.2 25.4 107156 52.6 47.4
花蓮縣 Hualian County 119669 73.0 22.8 151730 25.4 0.0
澎湖縣 Penghu County 36867 52.6 40.0 45910 49.4 48.1
基隆市 Jilong City 147613 65.0 30.7 156065 55.1 42.1
新竹市 Hsinchu City 162818 59.7 35.2 166563 55.6 41.3
嘉義市 Chiayi City 111827 49.1 47.1 134031 52.2 45.7
金門縣 Jinmen County 25979 90.7 4.6 38280 37.3 0.0
連江縣 Lianjiang County 4321 86.1 8.9 5482 98.6 0.0
total 3774338 57.1 37.8 4374932 47.9 45.3

That’s 26 races: nine by-elections and 17 county executive races.  In three of the county executive races, the DPP did not have a candidate.  In 22 of the other 23 races, the DPP increased its vote share.  (In Chiayi City, the DPP vote share declined by 1.4%.)  The smallest gains were 4-5%, but most of the gains were 10% or more.  8 races saw the DPP gain more than 15%.  You can tell a soap opera about each election and explain the outcome in purely local terms.  However, when the same thing happens in district after district, you have to start looking for national-level explanations.  In other words, the DPP’s popularity has significantly increased since early 2008.

How much has the KMT slipped and the DPP risen?  These 26 districts are slightly more favorable to the KMT than the national average.  In 2008, the national balance in the 2008 legislative party list was 55.2-40.4%.  Among the nine districts that had by-elections, the 2008 balance was 59.8-34.8%.  In the 17 counties that elected executives last December, the 2008 balance was 57.1-37.8.

It does not bode well for the KMT that it could only manage a 47.9-45.3% advantage in the county executive races.  A 20% edge in early 2008 was cut to less than 3% in late 2009.  If this 16.7% swing extended to the entire electorate, the DPP would hold a 1.9% advantage instead of a 14.8% disadvantage.

As bad as the county executive races look for the KMT, the by-elections look worse.  Originally, the KMT held 8 of the 9 seats.  It now holds 2.  Arguably, Yunlin 2 is a DPP-leaning district, but the other 7 seats looked like they were clearly more favorable to the KMT and the DPP won 5 of them.  Where the KMT had a whopping 25.0% advantage in 2008, it ended up with a 11.8% deficit in 2009-10, an unbelievable 36.8% swing.  These results don’t merely portend a KMT loss of power, they warn of a bigger DPP sweep than the KMT had in 2008.  If even districts like Taidong, Hualian, and Hsinchu are in play, then the KMT only has two safe seats left (Jinmen and Lianjiang).

Perhaps it is not that bad for the KMT.  Remember, the by-elections generally had around 40% turnout.  The exception is Nantou 1, which held its by-election on the same day as the county executive elections and thus had a “normal” 66.3% turnout.  Not coincidentally, the KMT won Nantou 1.  Its vote share slipped significantly, but not calamitously.  I argued above that the 20% of voters missing in a 40% turnout would disproportionately favor the demoralized party, in this case the KMT.  However, I also argued (with very little evidence) that the effect wouldn’t be too dramatic.  It would probably only sway the final vote share by a few percentage points.  With a 60% turnout, instead of 41.6-53.4%, we might get something like 45-50%.  That is still a disaster for the KMT, though perhaps not of the same magnitude.

There is another possibility.  Perhaps the DPP simply does better in by-elections.  The logic of this argument would go something like this.  KMT campaigns tend to be more organization-oriented, while the DPP campaigns tend to be more issue oriented.  With only a few election districts holding elections at any one time, the DPP can concentrate all of its rhetorical fire and oratorical superstars on a small number of voters.  At the same time, with all eyes on a few districts, the KMT finds it much harder to run a “traditional” (read: corrupt) campaign.  I am reluctant to bring up this argument since I don’t really believe it.  You can easily make the opposite argument about the KMT concentrating resources.  However, I am grasping for plausible explanations for implausible results in these by-elections, even given the 2009 county executive results.[4]

So where is the national partisan balance right now?  My best guess is that the two parties are roughly even.  Perhaps the DPP has a slight edge.

How could this have happened?  I have two answers, and I think both are basically correct.  One answer is that President Ma has frittered away his entire electoral advantage in less than two years.  By moving too quickly in opening to China, bungling the relief effort to Typhoon Morakot, and so on, he has alienated all the swing voters who supported him in early 2008.  TVBS has done a series of polls on party image.  The most recent was released on Jan. 20.  They ask a series of questions, including whether you agree or disagree that the KMT (DPP) pays attention to public opinion, is honest and not corrupt, is energetic, values reform, is united, has the ability to reflect on mistakes, and is trustworthy.  Opposition parties tend to do better on these questions, but during Chen Shui-bian’s term in office, it took a full four years before the KMT started to enjoy a clear advantage in party image.  The held this advantage for roughly the last three years of Chen’s tenure and the first few months of Ma’s tenure.  However, in less than two years, the trend lines have completely reversed, and now all these indicators are solidly in favor of the DPP.

A second answer is that the KMT never had such a huge advantage.  Ma’s landslide victory was not due to overwhelming public support for him or his party.  Rather, the electorate was disgusted with the scandals of the Chen administration and was punishing his entire party.  As the DPP steps away from the Chen era, we revert to a more realistic balance of power between the parties, one based on opinions toward the entire KMT and DPP rather than just toward the very polarizing Chen Shui-bian.  I don’t mean to say that it is suddenly 2003 among the electorate again.  Other interesting things have happened in the meantime.  Geographically, the biggest change in the electorate has been a clear shift in the Yunlin-Chiayi-Tainan region to the DPP.  I’ll almost certainly delve into this some other time.  However, the biggest story nationally is that this Chen era is over.  The DPP is a viable contender for power once again.


[1] Alas, this will not be an airtight argument.  This is simply my gut-feeling after years of attempting to come up with something better.  One of the advantages of a blog is that I can make these kinds of arguments.  This would never get past a referee.

[2] Why 60%?  60% is merely a convenient number since almost all general elections have gotten 60% or more, and I have not been able to figure out any relationship between turnout and partisan results in general elections.  Only by-elections have fallen significantly below 60%.  Over the past year, most by-elections have had roughly 40% turnout.

[3] The 2008 presidential vote is nearly identical.

[4] The exception is Taidong County.  Taidong produced nearly identical results in the county executive race and the by-election.  However, aborigines voted in the county executive race, and aborigines heavily support the KMT.  Subtracting them from the electorate in the by-election produces a narrow DPP victory instead of a narrow KMT victory.

Tags: , , ,

4 Responses to “by-election results”

  1. Michael Turton Says:

    Excellent post! Lots of good thinking.

    Bruce Jacobs argues that the bases are smaller than most people think and Taiwan voters are swing voters.

    The TVBS polls suck at predicting — they underestimate the DPP vote massively, I don’t know why. Probably because the Green people interviewed don’t want to tell the truth to an ardently pro-KMT station. But the TVBS poll also asks another question — :”Who do you think will win?” and for that one, the poll was dead right.

    The Taoyuan election was interesting, but I looked at the townships involved and three of four went DPP in the recent county chief election. The pro-KMT one was large, so the total difference was 3000 votes in favor of the DPP, which happened (coincidence) to be the margin of victory. Hsinchu to me was the big surprise — longtime Blue area. The DPP fielded a good candidate, the local KMT was divided, and the result was a blowout. Sweet!

    Hualien was never in play, DPP internals had Bi-khim down by 10% but closing, she lost by about 7.5%, as I recall.

    I agree with your analysis of Ma’s victory, and now that the smoke has blown over, buyer’s remorse is strong. A hidden factor in all this is the KMT’s idiotic decision to detain Chen. As long as he is in jail, silenced, Tsai can move the DPP swiftly into the post-Chen era, and he isn’t going to be out there stealing attention from the candidates, saying stupid bombastic things, and generally making people remember how badly he screwed up, as well as chasing away the swing voters, who hate him because they felt betrayed.

    Great post, lots of good stuff.

    Michael

  2. frozengarlic Says:

    Unfortunately for me, the DPP does not publish its internal polls. To the extent I have had access to them and been able to observe their polling procedures, I have been impressed with their accuracy and integrity. For obvious reasons, however, they choose to release the results only selectively, if at all.

    Your point about keeping Chen locked away is very sharp. I am not expert enough on the workings of the judicial system to make a judgment as to whether this was entirely a political decision made at the top levels of the KMT or whether the prosecutors have enough leeway to detain Chen on their own. Either way, you are right that it certainly has helped the DPP transition into a new era.

  3. Michael Turton Says:

    Are you busy on Thursday night? I’ll be in Taipei. Send me an email if you want to join a bunch of intelligent long-term expats here, and myself, for good conversation at a local burger joint.

    Michael

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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 )

Google+ photo

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

Connecting to %s


%d bloggers like this: