The state of the polls: 2019 vs 2015

Today is Dec 11, exactly one month before the election. Now seems like a good time to look back four years and compare the polls today with the polls four years ago. If you remember, Tsai Ing-wen eventually won the 2016 election with 56% to Eric Chu’s 31% and James Soong’s 13%.

During the summer and early fall this year, lots of people were speculating that the KMT would once again change its nominee. I argued that this was highly unlikely, in part (ok, mostly) because the KMT’s polling numbers were so much better this year than in 2015. This is no longer true. After Han’s recent slide in the polls, we are now in roughly the same place that we were in in Dec 2015. If anything, the KMT’s position might even be a little worse today than it was four years ago.


Here is my updated chart with the aggregated polling average for this election.

I didn’t do one of these four years ago, but fortunately our good friends who edit Wikipedia did a lot of wonderful work collecting polls and making pretty pictures. Here is their picture for the summer months when Hung Hsiu-chu was officially the nominee.

Tsai spent the first half of this period at about 40% and then drifted up to about 45%. Hung was in the low 20s in July and the high teens in August and September. Soong was actually slightly higher than Hung in August, which is one of the main reasons the KMT panicked and revoked her nomination. They were terrified of the possibility that they might slip into third place. As we are all quite aware, third place is not a good place to be in a majoritarian system like Taiwan’s. In mid-September, when the KMT announced the switch, Tsai’s lead over Hung was roughly 25-30 points.

Compare that to this year. In mid-September, when Gou and Ko had to make their decisions whether to run as an independent and the speculation about the KMT changing nominees was loudest, Tsai’s lead over Han was only about 10 points.

After the KMT switched to Chu, its polling numbers improved a little. Most importantly, Chu maintained a clear lead over Soong, about 10 points in October and shrinking to about 5 points in December. Maintaining second place was Chu’s most important duty, and he managed it adequately. During the last three months, Tsai held steady at about 45%, while Chu drifted downward from the low 20s to the high teens. By election day, Tsai led in the polls by about 25 points, which, not coincidentally, was exactly her margin of victory when the votes were counted.

This year, Tsai spent October and early November at around 45%. When Soong entered the race, her support dipped a few points. That is, her support this year was similar to her support levels four years ago. Han, in contrast, was considerably better than Chu, spending October and early November in the low 30s. The gap had grown slightly from September, but it was still only in the low teens.

Things changed dramatically in the second half of November, when news about Han’s Nangang luxury housing and the KMT party list dominated the political discussion. Since Soong announced his run on Nov 13, Tsai has gained about seven points and Han has lost about seven points. The gap between the two has exploded from 13% to 27%. This is firmly in the range that we saw four years ago.

You might be skeptical of the polls since Han asked his supporters not to participate or to say they support Tsai. Before, that took place on Nov 28, the polls had already begun to shift. The gap on Nov 28 was 21%, which is already arguably at the lower end of the 2015 range. I argued in a previous post that Han’s move probably didn’t have too much effect on the polling numbers. However, if you want to make a mental adjustment, feel free to choose a number somewhere between 21% and 27%. No matter what you choose, it’s a big number.

I think it is also notable that Tsai’s support is slightly higher this year than it was four years ago. In 2015, Tsai only broke 50% in two polls. This year, she has broken 50% in the three-way race in eleven polls, five of which were completed by Nov 28. (She also broke 50% in several two-way polls before Soong entered the race.) Since Soong is polling well below his 2015 levels, it is entirely plausible that Tsai could win by the same 25% margin while also beating her 2016 results by 3-5%, say 61%-36%-3%. Remember, the margin in the polls has historically been a pretty good predictor for the final vote margin, even though there are no undecided votes in the final tally. It doesn’t always work out exactly that way, but it is more likely than any other outcome.


This brings us to the second half of this topic: the legislative district races. I’m going to ignore the party list votes today because control of the legislature will be decided by the 73 single seat districts, which tend to swing disproportionately to one side or the other. The legislative races have historically been very similar to the presidential results. Again, there are sometimes individual legislative candidates who outperform the presidential candidate in their district, but the two vote totals are usually very close.

For the purposes of the legislative election, it is critical to remember that Tsai did not actually win the presidential race by 56% to 31%. Almost all Soong voters ended up voting for KMT district candidates, so the actual national partisan balance was 56-44%. That is, the green side’s margin of victory was 12%, not 25%. That 12% margin enabled the green side to win 53 of the 73 district seats.

In 2015, Soong polled pretty well. If you look at the poll numbers for Chu plus Soong, the blue side had around 30-35% in most polls. In the final polls, the blue camp trailed the green camp 44-33%.

This year, Soong is not polling as well, and the green camp lead in the presidential election is actually larger than it was four years ago. Right now, the green camp leads 49-30%. Even if you go back to Nov 28, the lead is 46-32%. That is, the lead is somewhere between 14 and 19 points. That might not sound too different from a lead of 11-12 points, but it is. Different seats have different tipping points, so each additional percent flips a new set of seats. Given the DPP’s fantastic performance in 2016, it seems almost unfathomable that they might not merely hold all those seats but actually win some new ones. However, that is what the polling suggests as a likely outcome.

At this point, I can almost hear everyone screaming at me. Yes, I know. Four years ago, most of Tsai’s supporters voted for DPP legislative candidates, but this year is different. The polls say that lots of Tsai supporters do not plan to vote for the DPP in legislative races. This might be correct. However, indulge me while I make two points.

First, I am not thinking about the party list vote. I expect that lots of Tsai supporters will vote third party on that ballot. Four years ago, the DPP got 44% of the list votes, 12% fewer than in the presidential race. (The KMT got 27%.) This year, the DPP might suffer an even bigger drop off from the presidential vote to the party list vote. You know what, I don’t much care. The party list vote is relatively unimportant. It takes about 2.5% to add or subtract a seat, so the losses would need to be massive before they affect the final seat total in a meaningful way.

Second, there are some polls that ask how people plan to vote in the district elections, but I don’t think these are very reliable indicators. I’m not sure how many people know who all the candidates are in their local race. Further, the two main blue and green candidates will finish first and second in all 73 races; there are no seats that promise to have anything resembling a true three-way race. As such, when it gets down to the end, if voters wish to cast a useful vote, they will have to decide which of the main two candidates to support. Minor candidates often find that their polling support evaporates on election day in the face of strategic voting. I think the presidential polls are probably a better indication of which side voters will eventually end up on in the legislative races.

But hey, just for fun, lets look at some polls. This year, MyFormosa has asked about the both the district and list races. In order to make sure Han’s ploy isn’t affecting things, lets look at the late November poll.

  president district list
DPP 51.2 33.9 34.1
KMT 23.7 22.9 25.1
TPP   4.5 9.5
NPP   2.4 5.3
PFP 5.2 0.6 2.5
Green   1.3 2.0
Statebuilding   1.1 1.7
others   1.3 2.0
Won’t vote 8.2 5.6 4.9
undecided 11.7 26.3 13.0

In this poll, there is a huge dropoff from the presidential vote to the district and list votes for the DPP, but no dropoff at all for the KMT. DPP support in the district races is only 65% of its support in the presidential election. One very reasonable explanation is that the KMT presidential vote has been stripped down to hardcore party loyalists, while the DPP presidential vote is an enormous (unwieldy) coalition of everyone else. Once all those people vote for Tsai in the presidential election, they will go their separate ways in the legislative elections. That’s reasonable. There are a couple of important features, though. For one, the number of undecided voters is twice as large in the district elections as in the list elections. That is a very big pool of floating voters, and, since Han’s voters are basically all accounted for, almost all of those undecided voters will have already expressed – however begrudgingly – a preference for the green camp over the blue camp. Second, even though the numbers of voters planning to vote for third parties in the districts is low, it is probably still too high. For example, 2.0% want to vote for the Green Party on the party list, and about 2/3 of them want to also vote for the Green Party in the districts. However, the Green Party is only fielding ten candidates. Probably fewer than 20% of Green Party list voters will have an opportunity to vote for a Green Party district candidate.

Ok, let’s see what the polls looked like four years ago. I’m going to look at the Dec 20, 2015 results; this was the last poll TVBS published which asked about both district and party list vote intentions.

  president district list
DPP 46 31 26
KMT 26 27 26
PFP 10 2 6
Independents   2 2
MKT   1 2
NPP   1 8
Green/SDP   1 2
TSU     2
New     2
others   1 2
undecided 17 33 20

Would you look at that! There is a huge dropoff for the DPP from the presidential election to the legislative races, and no dropoff for the KMT. DPP district support is only 67% of its presidential support. In the district elections, there is a large pool of undecided voters who mostly supported Tsai in the presidential race. This result looks eerily similar to the MyFormosa result from two weeks ago.

What didn’t happen four years ago? Even though this looks like a big tent coalition for Tsai in the presidential election that might splinter in all directions in the legislative races, that isn’t what happened. The legislative results, mirrored the presidential results to a striking degree. Will the same thing happen this year? There are no guarantees, but this reminder of what happened four years ago knocks away the strongest argument for expecting the DPP presidential coalition to split apart in the legislative races. Polls showing lower DPP support in legislative races do not necessarily actually portend lower support in the district races.


After staring at the polls for a long time, I’m coming to expect a substantial DPP victory in both the presidential and legislative races. It may be even larger than the 2016 victory.

Strangely, no one wants to hear this conclusion. I guess it’s reasonable that KMT supporters don’t want to hear it. It’s a nightmare outcome for them, and many would prefer not to face that possibility. (I must be strange; I want the bad news in advance so I can be psychologically prepared.) What is more unexpected is that none of my green friends are happy to hear my conclusions. They are terrified that something might go wrong if they are overconfident. If you talk openly about good things, you might curse it. They would prefer a tighter race so that everyone is a little scared and everyone feels the need to go out and vote. Well, I’m not trying to make you feel better. Or worse. I don’t care how you feel. My job is to give you my best answer, not the most politically correct answer. In fact, after staring at the data, I don’t think it is likely at all that the KMT will win a majority and I don’t think that it is likely that the DPP will fail to secure a majority. In fact, I don’t think it is likely that the DPP will lose more than a handful of seats. My best answer is that, as things stand right now, we are staring at another DPP tidal wave, roughly the same size and perhaps a bit bigger than the one four years ago.

10 Responses to “The state of the polls: 2019 vs 2015”

  1. Gold Says:

    So I have an ignorant question: Why was the KMT candidate winning 2nd place (ahead of Soong) such a big deal, as opposed to finishing 3rd behind Soong? (since you say that’s why the KMT panicked so hard and replaced Hung with Chu) I understand that it doesn’t look good for the KMT, in terms of appearance and saving face, but is there more to it?

    • Yuan Says:

      1) KMT used to be THE party in Taiwan. Like, the only party.

      2) PFP is a splintered off group from KMT, when President Lee chose Lien Chan to become KMT’s President nominee instead of Soong. Ah Bian won that 3 horse race (and coincidentally, KMT finished 3rd in the Presidential election that year).

      3) So why does finishing behind Soong/PFP matter that much when they have done it before? Because when Soong left KMT, he still had a huge support. Nowadays, he mostly stays out of the limelight, appearing in the news when Presidential election comes. After Soong’s 36% in 2000 (where Lien/KMT finished 3rd), he was VP for 2004, didn’t contest in 2008 and in 2012 only got 2.8%

      4) Should also take note that KMT and PFP generally has support from the same camp (pan blue voters). In a year where KMT wins, voteshare for PFP tend to be low. Voters seem to be voting for PFP when they can’t take KMT. Maybe.

  2. wei Says:

    the latest My-Formosa poll.

  3. Empty Says:

    Now that you’ve reached a well thought out conclusion, there’s not much left for you to do in your free (blogging?) time. May I suggest you consider + share your thoughts on all the possible black swan events that might dramatically shift the election landscape? It is not unreasonable, for example, to predict what might happen if:
    a) China “test-fires” a few more missiles over the Straits
    b) Han says something spectacular that moderate voters identify with
    c) Military conflict involving Korea/Japan/China/US etc.
    d) Soong/Guo/Ko get their act together to rally for KMT

    The possibilities are endless! Perhaps you might come up with even more plausible scenarios…

    Your blog is an amazing read – it really helps us all understand Taiwan.

  4. Quick Take: Frozen Garlic & legislative election - Taiwan Report Says:

    […] and… Courtney Donovan Smith42 mins ago14 min Today Frozen Garlic came out with a total home run of an analysis: This is the stuff that the Froze is absolutely brilliant at. Absolutely a must read.   […]

  5. squirrel Says:

    Frozen Garlic is my hero

  6. csempere109 Says:

    It relates to the strategic voting that Frozen Garlic alluded to. You might like a third party, but in this kind of electoral system you’ll probably decide that you’re better off voting for the bigger party on your side. If the KMT finished third to another pan-blue party, in this case the PFP, they’d be afraid that pan-blue voters would vote for PFP presidential and legislative candidates in future elections.

    Every cycle or two, it seems like people think the KMT or DPP are on the verge of collapse. Finishing behind a third party could make that narrative strong enough for it to actually happen.

  7. beidawei Says:

    I wonder which of the minor parties are likely to pass the threshold to enter the legislature. Just NPP and TPP? Is the PFP really moribund? (Wouldn’t be the first time–remember the Tai Lian Dang) I realize it won’t make much difference to the legislature, but it will have an effect on other would-be minor parties (like the one from two years ago with the yellow flags–I want to say Min Guo Dang–that didn’t win anything, and I guess just disappeared)

  8. watercress Says:

    Beidawei, it looks like Ko’s Mingzhongdang and the NPP are the only minor parties that have any real staying power in the upcoming decade or so, and even then I’m not sure about the NPP anymore. I think also that when Soong dies or retires, his orange party will become moribund.

  9. Radish Says:

    My only concern is the 3 upcoming presidential debates. Those are Han’s last chance to make up lost ground. If he lies convincingly and smoothly, enough times, he could gain a lot of votes.

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: