Archive for September, 2018

How have I never noticed this before?

September 21, 2018

I haven’t written anything for this blog in several months, so it is a little intimidating to try to start anew. After so much time away, I should probably be sitting on something really profound, something that will clarify or transform how you understand the upcoming election.


So instead, I’m going to write about something very trivial. This is something a stumbled upon a few days ago. Now that I’ve discovered it, I can’t believe I didn’t already know it. But again, in the big picture, it isn’t all that important.


In the 2016 presidential election, Taitung County had a turnout rate of 55.1%. This was somewhat lower than the national turnout rate (66.2%), which isn’t all that surprising. For people whose household registration is in Taitung but live elsewhere, getting back home is a considerable chore. You can’t just take the high speed rail, after all. For most people, a one-way trip is going to burn the greater part of a day. Travel is even more burdensome for people who don’t live near a major transportation hub. So it isn’t surprising that the turnout in Orchid Island, perhaps the hardest place to reach, was only 34.7%.

So far, there’s nothing unexpected. But humor me. For fun, let’s go observe the same pattern in 2014. Turnout in the county magistrate election (67.8%) was considerably higher than in the subsequent presidential election, which perhaps surprisingly, is almost exactly the same as the national turnout that day (67.6%). However, in Orchid Island the turnout was of course much lower, a mere 63.8%. Uh, hold on, that’s not very much lower. Something interesting is happening there.

I’ll be honest with you. I didn’t notice this pattern by looking at Orchid Island. I noticed it looking at county assembly elections in Pingtung. Pingtung has a whopping 16 county assembly districts, but 38 of the 55 seats are elected in the first four districts. I tend to focus my attention where the action is, so I usually don’t pay much attention to the other 12 districts, most of which are for indigenous voters. However, a few days ago I was looking at the last eight district, for mountain indigenous voters, and the turnout rates jumped out at me. Look at turnout in the following townships: 三地門 88.0%, 瑪家 86.7%, 泰武 88.9%, 來義 82.5%, 春日 87.2%, 獅子 85.7%, 牡丹 78.7%, 霧台 92.5%. You can go around the rest of the country and find the same pattern. Turnout in every mountain indigenous township is sky high, but only for local elections. In national elections, turnout is usually somewhat below average.

What’s going on? It’s not enough to say this is effective mobilization. Candidates all over Taiwan do their best to mobilize voters in every election, but turnout rates of 80% are extremely rare everywhere else. What these communities have that the rest of Taiwan lacks is, well, community (or what we call “social capital” in the academic literature). In these communities, everyone knows everyone, and they can use their social ties to make demands on one another. If your matriarch orders you to vote, you darn well better vote. If you don’t, word will definitely get back to her, and you are in for some stern looks. Likewise, if your neighbor (and everyone is a neighbor) is running, he will certainly ask you for your vote and will know if you didn’t vote. And if you don’t live in the village, your friends and family do, and they will reach out to you. It isn’t just negative penalties; working with your friends for a common goal is fun, even if you don’t really care about politics. These sorts of social sanctions simply aren’t available in the rest of Taiwan. Moreover, indigenous villages are bounded communities, and electing the town mayor is akin to electing a tribal leader. It’s a really big deal. (As to that 92.5% turnout in 霧台. I have a friend from Wutai. Wutai is populated by Rukai, a small tribe extremely active in trying to maintain its culture and identity, even in the face of having to abandon one its physical villages. Wutai’s social capital is through the roof.)

Interestingly, this local election mobilization effect doesn’t seem to work quite as well for plains indigenous voters. This might be because this group, which is overwhelmingly Amis, is larger and therefore not quite as socially cohesive as the mountain indigenous tribes. Or maybe it is that Amis tend to live in towns that are majority Han, so that the mayoral candidates are not usually one of their own. They get a bump, but it isn’t as large. Likewise, turnout is much lower for mountain indigenous voters who are registered in Han-majority townships. These voters might not be a bit more distant from the social networks, and they might not even be from the same tribe as the majority of the electorate. However, you can see the same effect in one specific Han society. Liuchiu Township in Pingtung is a small island with a clear identity somewhat differentiated from the more general Taiwanese identity. Turnout in the 2016 presidential election was 44.9%, but in the 2014 local election it was 78.7%. Social capital is powerful!

If you need more convincing for this social capital thesis, here’s one more little bit of evidence. I compared the turnout in 2009 for indigenous county assembly seats in Taitung with the turnout in the 2012 presidential election. After controlling for the overall different turnout levels, the turnouts in the local election were higher: 海端 +9.4%, 延平 +14.7%, 達人 +17.9%, 蘭嶼 +13.0%. However, there was one exception: 金峰 -7.5%. What?? It’s actually not hard to figure out. In Jinfeng that year, both the county assembly and town mayor elections were uncontested. With only one candidate who is sure to win, there is no need to activate those social capital networks. Sure, the county magistrate candidates tried to mobilize voters, but they can’t tap into the power of the social networks nearly as effectively. Grandma might mutter under her breath for weeks if you don’t vote for Second Cousin’s Wife, but she probably won’t take it personally if you are too busy to vote for #3 on the billboard (what’s his name again?).


I started this post by saying that this topic is trivial, and it is if you care primarily about who will win the city mayor and county magistrate elections. The number of indigenous voters is miniscule compared to the number of Han voters, so even a 20% bump in turnout for indigenous voters barely makes a ripple in the overall results. For example, there isn’t much chance that spectacular turnout among the 1800 or so indigenous voters in Wulai will be decisive in the New Taipei race. That result will be decided by the 3.2 million Han voters. In a different partisan balance, it might matter in Pingtung, where the astronomical turnout among the 40,000 mountain indigenous voters produces nearly 5000 extra votes (beyond presidential levels of turnout). However, Pingtung hasn’t had a close magistrate election since 1993, when Su Tseng-chang was beaten by 12000 votes. (I wonder whatever became of that guy…)

The only place where this local election mobilization bonus might impact the outcome this year is in Taitung. Some back of the envelope math suggests that the added turnout from local elections might be worth an extra 2000 votes from the 14000 mountain indigenous voters. I didn’t try to figure out how much the smaller bump from the 44000 plains indigenous voters might be, but let’s imagine that it is roughly also about 2000 votes.

We all know from years and years of experience that indigenous voters have tended to support the KMT over the DPP, so those extra 4000 votes are good news for the KMT. However, what none of us know is just how overwhelmingly indigenous voters support the KMT. If the KMT wins by a 95%-5% margin, then 4000 extra votes implies a net gain of 3600 votes for the KMT. Fantastic! If the KMT gets 80%, then they reap a net gain of 2400 votes. Still pretty good, though less likely to be decisive. If the KMT only wins 60%-40%, the net gain of 800 votes probably isn’t enough to matter at all. The thing is, you don’t know which of those scenarios is closest to reality. Your guess is as good as mine. (Out of curiosity, does anyone want to hazard a guess?) No one has ever put calculated a rigorous estimate of how indigenous voters vote in presidential elections, much less mayoral elections, so we just don’t know the answer.

Of course, that last sentence isn’t entirely true. It just so happens that I am working producing an estimate for indigenous votes. I’m just starting to put together results, so I don’t have a definitive answer just yet. However, I do have some interesting results from Taitung that I might write about if I ever get around to writing another Frozen Garlic post…