Presidential polls on the eve of the polling blackout

Today, Dec 31, is the last day that polls can be published before the ten-day blackout period starts. A couple last polls straggled in this morning, so I can now present the final weighted poll chart for this year’s election. (For methodology, please refer to the original post.)

This is an astonishing chart. All year long, Tsai Ing-wen’s fortunes have steadily improved while Han’s continually eroded. There was no single event that suddenly transformed the election; it happened bit by bit. While there was no single day that felt completely different from the day before, the race at the end of the year is completely different from at the beginning of the year. My chart only goes back to May, when we started to get a steady enough flow of polls that I could put together a daily average. However, if you look at individual pollsters with longer data series, you can see that the race had already started to change by May. In February, Tsai was much further behind. Here are the data series for the two most reputable pollsters this year, TVBS and Formosa. TVBS has a consistent blue bias, while Formosa tilts a bit toward the green side. Both show the same basic picture: Han starts out with a big lead, which slowly turns into a big deficit.  The “golden cross,” the point at which Tsai overtakes Han, is much earlier in the Formosa chart, but that is to be expected given the partisan skew of the two polls.

Do different types of people have different preferences? One problem with looking at demographic groups in normal polls is that the sample sizes are too small, so the numbers that pundits love to talk about (Han’s support in Hualien among women aged 30-39 has skyrocketed!!) are actually just random noise. In order to say anything reasonable, we need bigger sample sizes. Formosa conducted seven polls from mid-October through late December. Instead of looking at individual poll results, I have taken the average of these seven polls for various subgroups. This should yield a far more reliable look at the differences across categories.

Let’s start with sex. There has been a gender gap in Taiwan for decades. Men are more likely to support the DPP, while women are more likely to support the KMT. This gender gap surprises many outsiders, especially Americans, who assume that women should support the progressive side in Taiwan as they do in the USA. It doesn’t work that way, even with a woman leading the ticket. Tsai did better among men than among women both in 2012 and 2016, and the same thing will happen again in 2020.

We don’t know exactly why this gender gap exists. Most people think it reflects women’s desire for stability and safety, leading them to support the party promising it can avoid war with China. Now that questions concerning sexuality and personal autonomy, such as same-sex marriage and limiting abortions, as slowly entering Taiwanese politics, I’m curious to see if the gender gap will be affected, especially among younger or unmarried people.

Age has emerged as a very important variable in the last decade. As recently as 2012, there was not much of an age difference. Now, the gap is dramatic. Among people over 40, Tsai leads Han by 15 to 20 points. Among the 30-39 group, her lead is nearly 30 points. Among the 20 somethings, it is 45 points.

Younger people have strong Taiwanese identities and overwhelmingly dislike the KMT, but that does not imply that they love the DPP. In fact, they have decidedly mixed feeling toward the DPP. They will vote for it if the individual candidate is palatable (as with Tsai), but many would prefer to vote for another party. In non-plurality elections, younger people often look elsewhere.

Tsai leads all education levels by roughly similar amounts. A glance at 2016 pre-election surveys shows similar patterns. Nothing to see here, folks. Move along.

Region is interesting. Tsai is strongest in the Tainan and Kaohsiung areas, while Han is stronger in Taipei, New Taipei, and Taoyuan. We all know that Taiwan is blue north, green south, so this makes perfect sense.

Uh, wait a second. Did you notice that Tsai is a lot more popular in New Taipei than in Taichung, which are traditionally pretty similar for the DPP. In fact, Tsai is nearly as strong in Taipei and Taoyuan as in Taichung, which is not historically the case.

Let’s directly compare those polling averages to Tsai’s actual votes in 2016. The gap shows that she is doing a bit better in New Taipei and a bit worse in the Taichung area. People like to say that central Taiwan is the traditional bellwether zone; if you win it you will win all of Taiwan. That is correct, but it’s more accurate to expand the bellwether zone to central Taiwan and New Taipei, which historically fall at just about the same point on the partisan divide. However, this year, New Taipei is considerably greener than the Taichung area.

The differences between this year’s polls and the 2016 results are intriguing on the green side, but they are downright stark on the blue side. Compared to Eric Chu, Han is reasonably close to Chu in central, southern, and eastern Taiwan. The Tainan area may be his weakest region in absolute terms, but he is actually polling quite well there compared to the KMT’s past performance. However, northern Taiwan, with nearly half the population, is a disaster. The problem for the KMT is that, unless Han outperforms his polls by a stupendous amount, the swing legislative districts are not likely to be in the south. The DPP should have enough of an advantage there that it can absorb a bit more Han popularity with no effect on the legislative races. However, the north is full of close races, most of which tipped in the DPP’s direction last time. These numbers suggest that it will be very hard for the KMT to win many of those back. Many races in central Taiwan were close four years ago, and this is the only region in which these geographical shifts might help the KMT. Unfortunately for the KMT, there are 30 seats between Hsinchu City and Keelung City in the north, compared to only 14 in central Taiwan.

One key to Tsai’s resurgence is that her approval rating has improved immensely since last year. For the first three years of her presidency, her approval rating experience a long, sad, steady slide. By last year’s local elections, according to many polls, her approval rating was near a paltry 20%. That is not a good way to go to the voters to ask for renewed support, and her party was flogged severely. However, over the course of 2019, Tsai’s approval rating has recovered dramatically. From being over 40% underwater, Tsai now has a net positive approval rating in most polls. This chart is Formosa’s tracking poll, which they have compiled over her entire presidency. I don’t know exactly what the numbers in Nov 2018 — not only the election but also her worst month in this chart — were, but the numbers for Dec 2018 — which was nearly as bad — were 21.5% approve and 67.3% disapprove. Ouch.

Tsai’s political resurgence was not merely a personal triumph. She, her party, and the entire green side of the political spectrum have risen together. Look at trends in party identification. Party ID is measured by a question asking respondents which party they generally support. Party ID has a very close relationship with vote choice, and it is considered one of the most important variables in voting research. The Formosa polls break down Party ID by individual party and also aggregate those results into blue and green camps. In practice, very few people identify with small parties, so blue party ID is almost all KMT party ID and green party ID is almost all DPP party ID. (Voters may support smaller parties, but they are less likely to have a strong, enduring psychological attachment to a particular small party, many of which are relatively new and unknown.) This chart shows blue and green party ID over the past year. There is a gradual increase in the green camp’s popularity from February to November, when it suddenly shoots up considerably. The blue camp’s popularity showed a gradual decline over the entire year. As a result, the green camp went from facing a 7.7% deficit in party ID to enjoying a 17.4% advantage.

Another way to look at this is by breaking the electorate down into different types of groups, based on how they think about the two big parties. The Formosa polls put people into nine different categories, three who prefer the KMT to the DPP, three who prefer the DPP to the KMT, and three who are neutral. You can arrange the KMT and DPP categories spatially, from left to right. However, the three neutral categories don’t have a specific spatial ordering. Some people have positive feelings toward both parties, some people have negative feelings toward both parties, and some people are uninformed and can’t make judgments about either party. These three groups behave in quite different ways. (I described these nine groups in more detail in an earlier post.)

If you look at the distribution of these nine groups over the past year, you can see that the three pro-DPP categories make up a bigger and bigger part of the electorate while the three blue categories have shrunk. If you take the average of the first three surveys (Feb-Apr), 36.3% of the electorate was blue while only 26.4% was green. In the last three surveys, only 27.3% was blue, while a whopping 42.1% were classified as green. A 9.9% disadvantage for the green side early in the year had transformed into a 14.8% advantage late in the year.

We naturally expect the DPP candidate to do well in the pro-DPP groups and poorly in the pro-KMT groups. That is exactly what happened. Tsai only did markedly better in one group, the neutral voters who like both parties equally. With all the other groups, she won roughly the same percentages both early and late in the year. Her gains came almost entirely from expanding the pro-DPP groups rather than by becoming more personally popular.

Han is a different story. He lost some support because his party declined in popularity. However, his slide can’t simply be blamed on the KMT’s collective brand. Compared to his early 2019 polls, Han became significantly less popular within each group. That is, he won less support among die-hard KMT supporters, KMT leaners, people who liked both parties equally, people who dislike both parties equally, and all stripes of pro-DPP voters. He bears a heavy personal responsibility for his failure in the polls.

Since I know you are curious, Soong gets a little bit of support from all nine groups. His best group, not surprisingly, is voters who don’t like either big party. They might like Soong and the PFP, or they might just be casting protest votes against the two big parties. It is worth noting that Soong doesn’t do nearly as well in this category as Ko Wen-je did. Back in the spring and summer, when many people thought Ko was running, he would usually get 50% or more in this category.

 

Well, that’s it folks. No more polling until Jan 12. Enjoy all the groundless speculation about underground gambling odds.

27 Responses to “Presidential polls on the eve of the polling blackout”

  1. Shelley Rigger Says:

    Thank you for the prodigious and important work that I, for one, have been freeloading off of for months. What would we do without you? Happy New Year, Dr. and Mrs. Garlic!

  2. Rue Says:

    You are the GREATEST source. I was waiting for just something like this

  3. Green Says:

    Hi Garlic, a few questions that have been on my mind, curious about your take on it:

    1. A Japanese diplomat (who accurately predicted the outcome of the 2018 elections) claimed that Tsai would only win by 5% and the race is much narrower than polls show, what do you think?

    2. Are you going to do a write-up on the Anti-Infiltration Bill? (it successfully passed the LY by a vote of 67-0)

    3. The International News Lens’ poll aggregate showed Tsai’s lead shrinking abruptly from 31% to 18%, is it an outlier? (it doesn’t show on its graph, but they suddenly published new numbers that replaced their old numbers)

  4. Phil Saunders Says:

    Great info to try to sort out the campaign and (soon) the election.

  5. Guy Beauregard Says:

    Epic post! Thank you!

  6. Mike Says:

    Fascinating insights. Was just wondering, for the plots comparing the regional support with the data from the 2016 election, would it be worth considering a comparison of the polling average (民調支持度) in each region with the 2016 election results, but looking at the vote shares as a percentage of the total number of eligible voters (e.g. multiplying voter turnout by vote share, or 投票率 x 得票率; or 得票數÷總選民人數) instead of considering vote shares alone (得票率)? Would the incorporation of voter turnout (投票率 x 得票率) potentially provide a little further insight into how the regional variation in vote shares might pan out in the 2020 election? (Quite curious to see what Soong’s regional distributions are as well, and how it might have shifted from 2016 to 2020 as well).

  7. Carrington Bowen DavisD Says:

    Did you actually say “a couple last polls”? No, you’re a skilled, trained journalist with a major medium and you care about communications with your audiences and are an educated communicator with a strong vocabulary and sense of appreciation for the English language and you would never speak or write broken colloquial English and mislead people with country-street talk like next couple times or last couple weeks because that’s not EVEN English, means nothing and says nothing. It’s irritatingly incorrect and improper so stop it the next couple times.( That should be the next couple of polls if you can’t find a dictionary or Thesaurus.

  8. Carrington Bowen Davis Says:

    You could have said the next polls or upcoming polls and be clearer and less misunderstood if you had left out the indefinite small sum work “couple” or the correct t vernacular “copula” which is equally a wasted word. Couple times!!!

  9. Meng Says:

    Something puzzles me about that gender gap chart. If Tsai is really ahead of Han by 20-30 points, shouldn’t both men and women be favoring her by a significant margin? It should look quite different than what the yellow and purple bars show.

    Also, does this all still take the “Han asked his voters to lie to pollsters” into account?

  10. Ginger Says:

    My worry, now, is that the helicopter crash may give the KMT a boost and drop the DPP.

  11. The Limits of Gender Bias in Perceptions of Candidates – Taiwan Insight Says:

    […] of her gender or the gender of the respondents. In addition, in contrast to patterns elsewhere, men have been more supportive than women of the DPP in general and of Tsai Ing-wen in particular, although it is unclear to what extent Tsai has created coattail effects for female legislative […]

  12. Ben Cheney Says:

    The Han rally in Taipei this evening is huge and everyone attending will vote 😬

    • Chiang John Says:

      Remember though, those are his hardcore fans. It looks impressive till you realize that’s basically his entire base. Middle aged to elderly.

      50,000 hardcore fans looks impressive, but what wins at the end is numbers. 100,000 barely interested voters is much more.

  13. The Guardian view on Taiwan’s election: an extraordinary comeback – then what? | Editorial – Piskavica Says:

    […] so dim that senior party figures urged her not to run again. Yet for months, polls have given her a hefty lead over her main […]

  14. The Guardian view on Taiwan’s election: an extraordinary comeback – then what? – The American Investor Daily Says:

    […] so dim that senior party figures urged her not to run again. Yet for months, polls have given her a hefty lead over her main […]

  15. What to Know About Taiwan’s Presidential Election – Knowing Everything Says:

    […] the past eight months have seen a long, steady slide in Mr. Han’s popularity. He is facing a recall drive from Kaohsiung residents who criticized him for running for president […]

  16. What to Know About Taiwan’s Presidential Election | Media One Says:

    […] the past eight months have seen a long, steady slide in Mr. Han’s popularity. He is facing a recall drive from Kaohsiung residents who criticized him for running for president […]

  17. What to Know About Taiwan’s Presidential Election – Whole W News Says:

    […] the past eight months have seen a long, steady slide in Mr. Han’s popularity. He is facing a recall drive from Kaohsiung residents who criticized him for running for president […]

  18. What to Know About Taiwan's Presidential Election | csbnnews Says:

    […] the past eight months have seen a long, steady slide in Mr. Han’s popularity. He is facing a recall drive from Kaohsiung residents who criticized him for running for president […]

  19. What to Know About Taiwan’s Presidential Election – The American Investor Daily Says:

    […] the past eight months have seen a long, steady slide in Mr. Han’s popularity. He is facing a recall drive from Kaohsiung residents who criticized him for running for president […]

  20. What to Know About Taiwan’s Presidential Election Says:

    […] the past eight months have seen a long, steady slide in Mr. Han’s popularity. He is facing a recall drive from Kaohsiung residents who criticized him for running for president […]

  21. Taiwan’s Election is a Vote About China - Diplomacy24.com Says:

    […] But these measures have proved ineffective so far: Ms. Tsai appears to have garnered considerable support from young voters in the current presidential […]

  22. Opinion | Taiwan’s Election is a Vote About China Says:

    […] But these measures have proved ineffective so far: Ms. Tsai appears to have garnered considerable support from young voters in the current presidential […]

  23. The Rise and Rapid Fall of Han Kuo-yu » News of The Secret inside the LifeStyles Says:

    […] announcing his presidential candidacy, early polls showed Han with a nearly 30-point lead over incumbent President Tsai Ing-wen of the DPP. But eventually, […]

  24. The Rise and Rapid Fall of Han Kuo-yu – The Diplomat Says:

    […] announcing his presidential candidacy, early polls showed Han with a nearly 30-point lead over incumbent President Tsai Ing-wen of the DPP. But eventually, […]

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 )

Google photo

You are commenting using your Google 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: