I’ve spent most of the past week digging through mountains of data from the Taiwan Election and Democratization Surveys (TEDS) trying to put together a paper proposal for a conference later this year. As a side effect, I have lots of stuff to share on my blog.
After last year’s elections, I lamented that we would never be able to completely figure out what happened in the two most important elections, New Taipei and Taoyuan, since TEDS was doing the big post-election face-to-face surveys in Taipei, Taichung, and Kaohsiung. Happily, I was wrong. In addition to the major surveys (which will be released in the next few weeks), TEDS also did pre-election telephone surveys in New Taipei, Taoyuan, Yilan, and Yunlin. Even better, TEDS has conducted national surveys quarterly since September 2012. As a result, there is a lot of stuff to dig through, and I might be able to come up with a more complete answer for why the KMT lost Taoyuan and barely won New Taipei.
Blue supporters are mostly ignoring last year’s elections. They don’t matter. They were local, not national elections. People just wanted to express dissatisfaction with President Ma, but they’ll come back to the KMT in national elections when it really matters. The KMT had lousy candidates. Whatever the reason, I keep talking to KMT true believers who think the KMT is in good shape for next year’s elections. They aren’t convinced that Hung Hsiu-chu can’t beat Tsai Ing-wen, to say nothing of the possibility that the KMT will lose the legislature.
Those objections are a little correct, but they are mostly wrong. Local elections are a bit different, but mayoral elections still run largely along party lines. The bigger the city or county, the more nationalized the election is. Hualien and Hsinchu County had weird, local things happen, but that type of thing is a lot less likely in a direct municipality. Sean Lien was a historically awful candidate in Taipei City, and he managed to single-handedly lose that race. However, the KMT candidates in Taoyuan and New Taipei were both more highly rated than their DPP opponents. Candidate quality can’t explain the poor KMT performance in those races.
Then there is party ID, which is what I’m really going to write about today. To put it bluntly, the KMT has suffered a massive decline in its party ID over the last four years, and party ID is one of the most important variables in all of political science. You can see this decline in data from TISR and the Election Study Center, NCCU, pictured below. From the late 1990s until 2012, party ID was fairly stable. The blue camp, mostly the KMT, had a consistent lead of about 5-10 points over the green camp, mostly the DPP. Not coincidentally, the blue camp consistently had about a 10% edge in most elections. In hindsight, the 2012 election might be both the most “typical” election result and also the last election of that party system.
A quick review. Party identification has two classic conceptualizations. The social psychologists of the Michigan School thought of party ID as a group identity. Someone would identify themselves as a Democrat in the same way they would identify themselves as a Catholic, a German, a Red Sox fan, or a union member. All of those identities define who the person is, so Democratic identifiers usually vote for Democratic candidates because they are both part of the same meaningful group. A person who ceases to identify as a Democrat is telling you something very substantive and meaningful about how he or she has changed. The other way to conceptualize party ID is as a running tally. This idea has its roots in the rational choice school of thought that comes out of microeconomics. According to this school, every time something happens, a voter updates his or her current opinion of the party. If something negative happens, the voter’s opinion is lowered. This running tally is then a summary of how the voter currently sees the party, and it is a good information shortcut to use in the voting decision. In Taiwan, party ID is usually operationalized as asking the voter, among parties A, B, C, D, and E, which party do you support more? A long list of studies over the past twenty-five years have shown that party ID is a powerful indicator of vote choice in Taiwan, just as in the rest of the world.
Here is the TEDS party ID data for the past four years.
The first data point is from rolling telephone surveys in the five weeks before the 2012 election. The second data point is from the post-election face-to-face survey, which was mostly conducted during the month after the election. The remaining data points are the quarterly telephone surveys. The surveys before and after the 2012 had large samples (n~5000, 2000), which the quarterly surveys had about 1000 interviews each. In some of the following graphs in which the data are cut into several categories, the quarterly data will jump around a bit more, reflecting the larger sampling error. The DPP held steady at around 25% through most of the period, but it has been above 30% in the two most recent quarters. Of course, the December 2014 data are critical, since they were taken right after the election. The KMT data is more dramatic. KMT party ID had a spike up from its normal 35% or so right before and after the 2012 election. By the time the quarterly data start in Sept 2009, this spike is completely gone. The KMT continues to bleed support, with a noticeable plunge in Dec 2014. Comparing the two elections, the KMT crashed from 43% in late 2011 to 23% in late 2014.
What’s amazing to me about this plunge is how it happens in nearly every sub-population. Maybe you think young people are abandoning the KMT. They are, but not any faster than old people. (I ran a binary logistic regression model on this for the Sept 2009 to Dec 2014 period, and the slopes of the individual lines are not statistically different from the slope of the overall line.)
Education isn’t the answer. All these lines go downward at just about the same rate. (Region and gender don’t show any differences either, but I’ll spare you those charts.)
Occupation is not quite uniform. KMT support among government employees (the blue line) declines at a slightly steeper slope. The red line for students is just about at the average until the June 2014 survey, when it plunges dramatically. It is as if a generation of students were radicalized or something! Statistically speaking, my model showed that the slope of the student’s line was more negative than that for the government employees. However, since students are a small group, their coefficient was not statistically significant.
There is a clear trend in ethnic background. Support for the KMT declined much less rapidly among Hakkas than among Mainlanders or Min-nan respondents. (I wish the sample sizes were large enough to analyze Aborigines, since there are hints of massive changes from the electoral returns.)
There is one more demographic variable that I find intriguing. I recoded all the townships into four categories. The first is the “urban core.” This includes all the prosperous parts of the major cities. The second is the “urban sprawl.” This includes the decaying downtown sections as well as the new growth overflow suburbs. Most of New Taipei and Taoyuan are in this second category. If money were no object, almost everyone would choose to live in the posh first category rather than the (comparatively) low-rent second category. The third category includes rural Min-nan townships. This category is dominated by the stretch of townships in the rural south from Changhua to Pingtung. The fourth category is much smaller and includes all the other rural townships. This group is dominated by predominantly Hakka townships, though it also includes a large number of (sparely populated) Aboriginal townships. There lines are different, especially if you limit the sample to the period from Sept 2012 to Dec 2014, as my model did. Support for the KMT among people in the rural diverse townships did not decline much at all. This is similar to the trend among Hakkas that we saw above, but it is even stronger here. It is possible that preferences among rural Hakkas have been more stable than those among urban Hakkas (though I haven’t tested that idea). The bad news for the KMT is that their best group is by far the smallest. The largest category is group 2, the urban sprawl. In this group, support for the KMT plunged the fastest. It’s hard to see in this picture, but the difference is statistically significant. TEDS telephone surveys don’t ask respondents for income information since that is too sensitive to do on the phone, but an obvious interpretation is that poorer urbanites are abandoning the KMT ship faster. This might be evidence of the emerging class cleavage.
The variations among subgroups are interesting, but the main takeaway point from this post is the main trend. Those big, black lines in the middle of each graph are moving relentlessly downward. The KMT can tell itself that this doesn’t matter. All those newly undecided voters will come back to the KMT when national power is at stake. That’s what the DPP told itself in 2007. That didn’t work out so well for the DPP, and the dip in DPP party ID in Chen’s second term was much smaller than the dip in KMT party ID during Ma’s second term. Whether people are no longer expressing a group identity with the KMT or their running tallies no longer put the KMT in a favorable position, this drop in KMT party ID is almost certainly the main cause of the KMT’s 2014 debacle (outside of Taipei). Unless things turn around in a big way, it is also almost certain to have a major impact six months from now.