What happened in Taoyuan?

So what the hell happened in Taoyuan? I didn’t expect John Wu 吳志揚 to lose, and no one I have talked to saw it coming either. I went to Cheng Wen-tsan’s 鄭文燦 rally on two nights before the election, and I don’t think anyone there expected it either. It was a rather small event. They had extra seats and a big grassy area with an extra screen just in case more people showed up, but they didn’t even fill the main seating area. I guessed that about 4000-5000 people were there. The atmosphere wasn’t exactly electric. The true believers were there fighting the good fight, but knowing in their hearts that the real battle would have to wait until 2016. When Tsai Ing-wen told the crowd that Cheng would win and asked them (as they do at every rally for every hopeless candidate), “Do you have confidence?” The answer “Yes!” was decidedly halfhearted.

Here, the Light Bulb makes a point. He is really a master on the stump.

Maybe this is when Tsai asked the crowd who would win.

Lots of empty spaces where there were supposed to be more people.

Then on Saturday night, we all watched in disbelief as Cheng took an early lead in the counting. A half an hour later, it dawned on me that we were no longer merely looking at a few strange early reporting ballot boxes, and it was possible that Cheng might win. When the DPP finally lost its lead in New Taipei but Cheng expanded his lead in Taoyuan, my brain nearly went into shock. Rationally, I could see what was happening and tell you that Cheng was going to win, but that simply didn’t make any sense to me.

Let me start by saying that we probably won’t ever be able to decipher exactly what happened because we don’t have data. Taiwan doesn’t allow exit polls, so we will never precisely know the demographics of voters who turned out or stayed at home. The main academic post-election survey won’t cover Taoyuan. Due to budgetary constraints, we simply can’t do a survey in every jurisdiction for local elections. This year, the Taiwan Election and Democratization Survey (TEDS) will study Taipei, Taichung, and Kaohsiung. A week ago, those seemed like a fairly obvious choices. (They are also the ones studied in the last election round, and if you want to look at continuity you can’t change places every cycle.) There will, of course, be telephone surveys, but you can ask far more questions (and far more complicated questions) in face-to-face surveys. Moreover, post-election surveys have the same limitations that pre-election surveys had. If we weren’t reaching the people who made up this anti-KMT wave before the election, how can we be confident we are reaching them after the election? At any rate, the point is that this election result will probably always be something of a mystery.

(You can also believe that losing candidates will use it for the next decade to try to inspire their supporters. “Who cares about the polls? Remember how the polls said Cheng Wen-tsan had no chance? We’re gonna win!”

Still, maybe we can find something in the patterns of votes.

My first reaction was that this result had something to do with ethnic politics. Taoyuan is often divided into a Min-nan north and a Hakka south. A few decades ago, there was some discussion about whether the county should be split into two new counties. That never happened, but the KMT incorporated the ethnic divide into its unofficial rules. The county magistrate always rotated between Min-nan and Hakka. In Legislative Yuan and Provincial Assembly nominations, the KMT was always careful to nominate the appropriate number of Hakka and Min-nan candidates. Those candidates were then expected to campaign in their half of the county, so that Min-nan candidates got almost all of their votes in the north and Hakka candidates were concentrated in the south.

The traditional Min-Ke divide was further complicated by the Mainlander population. Taoyuan is also home to an enormous military presence. Unlike Taipei’s Mainlander population, which was much more diverse (in terms of social status, education, occupation, and income) and was spread out into the regular society, Taoyuan’s Mainlander population was much more concentrated in military villages. The townships with the largest military populations were Bade (north), Longtan (south), Zhongli (south), and Pingzheng (south). In the last two decades, the military has razed almost all of the old military villages. Some of the previous residents are housed in newer military-built buildings, but many have moved out into normal society. The residents of the newer buildings are also not exclusively military families. The units can be bought and sold, so regular people can also move in. As a result, the old segregation of military families from regular society is no longer so prevalent. Still, Taoyuan County – especially the four aforementioned townships – continues to have a disproportionate number of Mainlanders.

Further muddying the picture is that Taoyuan has experienced tremendous population growth over the past three decades. This has diluted the traditional ethnic residence patterns quite a bit. It also continually brings in new voters who are outside the traditional mobilization networks.

Still, this race featured a Min-nan candidate (Cheng) against a Hakka candidate (Wu), so you would expect to see Cheng doing particularly well in the north. In fact, Cheng beat Wu by a margin of 54.6 – 47.2 in the seven northern townships, while Wu beat Cheng 51.7 – 44.4 in the six southern townships. Aha! It’s an ethnic war! Not so fast. For one thing, the same pattern held five years ago, when Wu barely Cheng in the north (49.7-48.7) but swamped him in the south (54.9-42.6). Cheng went up everywhere, not just in the north. Moreover, the north-south gap is baked into the party system. The DPP always does better in the north than in the south. For, example in 2012, Tsai Ing-wen got 41.6 in the north, but only 37.4 in the south. The gap was a bit wider in this election, but there isn’t much evidence for the idea that Cheng won with an unprecedented mobilization of Min-nan voters.

A second idea is that turnout was decisive. Lots of people are comparing the raw numbers of votes with previous elections an claiming that they can see evidence that blue voters stayed home, new voters turned out in droves, or something else. I’ve stared at the data for a few hours today, and I just don’t see it.

The CEC hasn’t released the full file of official turnout data yet. I can approximate the turnout at township levels by looking at the number of eligible voters (which they published a few weeks ago) and adding up the number of valid votes for each candidate. This is only an approximation, since it does not include invalid votes. In this case, it also might be fruitful to look at neighborhood-level data, but that would take a lot of time and we don’t have that data yet anyway.

		2009	2014	Increase	2009	2014	Increase
		Cheng	Cheng	Cheng	Turnout	Turnout	Turnout
桃園	Taoyuan (n)	48.4	56.3	7.8	50.1	57.9	7.8
八德	Bade (n)	45.2	50.6	5.4	52.0	62.4	10.4
蘆竹	Luzhu (n)	52.5	57.4	4.9	53.8	59.7	5.9
大溪	Daxi (n)	51.2	55.3	4.1	60.8	63.2	2.4
大園	Dayuan (n)	53.2	56.6	3.4	61.7	65.7	4.0
龜山	Guishan (n)	47.3	53.8	6.5	50.2	58.2	7.9
復興	Fuxing (n)	22.3	27.2	4.9	57.7	74.5	16.8
中壢	Zhongli (s)	39.5	45.6	6.1	54.5	62.2	7.7
平鎮	Pingzhen (s)	41.9	48.1	6.2	51.3	62.5	11.2
楊梅	Yangmei (s)	42.2	45.6	3.4	53.3	61.6	8.3
龍潭	Longtan (s)	38.6	44.9	6.4	54.3	61.1	6.8
新屋	Xinwu (s)	55.0	53.6	-1.4	66.1	72.2	6.2
觀音	Guanyin (s)	56.9	55.3	-1.6	61.3	69.4	8.1
	.
	total	45.7	51.0	5.3	53.7	61.5	7.8

One idea is that hard-core deep blue voters were disgusted with Ma’s performance and simply stayed at home. The overall (estimated) turnout in Taoyuan was rather low, at 61.5%, so many people did not vote. However, it was even lower in 2009 (53.7) when the KMT won. Looking at the township level, if the deep blue voters didn’t turn out, the effect should be biggest in the places the KMT is strongest. Alternatively, it could be the opposite. Perhaps only the deeply committed KMT voters turned out, and the less committed stayed at home. (Ignore Fuxing. It is an aboriginal township, and the patterns may be completely different. It also is tiny, so it can’t be driving the overall patterns.) Cheng is weakest in Zhongli, Pingzhen, Yangmei, and Longtan. However, the turnout increase is just about at the overall average in three of these, and it is a bit higher in Pingzhen. Cheng’s support increased a bit more than average in three of them, but in increase quite a bit below average in the other (more on that later). There doesn’t seem to be support for this hypothesis (or its opposite).

The Mainlander/military presence also doesn’t seem to be driving things either. Three of the four military townships overlap with the strong KMT townships. Trading Bade for Yangmei doesn’t produce any clear patterns.

Maybe someone with sharper eyes than mine can find something in the turnout numbers, but I just don’t see anything. This is not a surprise to me. In the days before every election, pundits scream that turnout will be decisive. They make amazingly precise predictions. If turnout drops below 65%, the DPP will win for sure. If it is over 68%, the KMT will definitely win. Or maybe it’s the other way around. Everyone has a guess about how turnout will affect the race, but they often have very different stories and come up with very different conclusions. The problem is that all these stories require heroic assumptions about what kind of people are ambivalent about voting, and there is no evidence for these assumptions. Many of them are based on nothing more than wishful thinking. After two decades of staring at data, I have never been able to find any consistent effect of turnout.

I do see one interesting pattern in the election results. Comparing Cheng’s vote share in 2009 and 2014, the two southern coastal townships stand out. Cheng actually got more votes in these two townships in 2009 than in 2014. Moreover, the next smallest increases are in neighboring Dayuan and Yangmei townships. Something about Xinwu and Guanyin townships is different, and that may be the key to understanding what happened in Taoyuan this year.

Two possibilities come to mind. First, the coastal areas are traditionally the DPP’s strongest areas. Second, the southern coast is the least industrialized part of Taoyuan. In my previous post, I pointed out that the national wave was larger in northern and central Taiwan and smaller in the rural parts of southern Taiwan. Here we are seeing the same thing. Just to corroborate this point, the next smallest increase is in Daxi, which is another corner of Taoyuan relatively less affected by industrialization and fast population growth and which traditionally is one of the DPP’s better areas. The biggest increases are in Taoyuan and Guishan townships, where much of the recent population boom has been centered.

It is going far beyond the data to speculate why we are seeing this pattern. It could be anger over food safety, which might be a non-partisan middle class issue. It might be something about urban class conflict. It might have something to do with the Sunflowers. However, I’m becoming more and more certain that the answer has to have something to do with city life.