TAICHUNG (Taiwan News) — Poll after poll shows Democratic Progressive Party (DPP) candidate Lai Ching-te (賴清德) in the lead.
Including polls taken through the middle of the day on New Year’s Day, the data inputted into the weighted Taiwan News Poll of Polls that tracks polling from the previous 15 days has Lai leading Kuomintang (KMT) candidate Hou Yu-ih (侯友宜) by 6.64% and Taiwan People's Party (TPP) candidate Ko Wen-je (柯文哲) by 11.29%.
The respected Formosa poll now has Lai leading Hou by 11.1% and Ko by just under 20%. The outlet predicts based on its polling that Lai will win between 40.6% and 41.8%, Hou between 36.7% and 38.5%, and Ko ending up with between 20.6% and 21.7%.
When I began this column, Lai was leading Hou by 4.13% and Ko had been sliding in the polls. Much of an entire day was spent preparing research and number crunching to estimate what percentage of the undecided voters and Ko supporters might consider switching their vote.
About a week ago, I estimated about 16.6% (realistically 15-18%) was the absolute maximum percentage that could be poached by one of the leading candidates. This included the estimated turnout of undecided people and the maximum percentage of Ko supporters who might be willing to strategically switch their vote to their second favorite candidate because they decided Ko had no chance.
Things looked good for Lai because Hou would have had to win over three quarters of those voters, and there were indications that if Ko supporters switched their vote, possibly over half would go to Lai. In the Formosa poll at the time, 41.7% of Ko supporters said they intended to vote but did not count as an “iron vote” (鐵票, voters who will support their candidate no matter what), which meant, of the 16.6% estimate, 8.9% was the absolute maximum that might come from Ko supporters.
Ko's supporters will stand by him
What a difference a week makes. I was wasting my time. I think the number of Ko supporters who will strategically switch their votes will be very low.
Of the ten polls counted in the poll of polls in the last 15 days, three show Ko ahead of Hou, three show Hou crushing Ko by around ten points, and the rest have Ko behind Hou, but not by large numbers. Ko’s supporters are younger and largely get their news via their social media networks, and the TPP has been more adept than the other parties at utilizing this.
This means Ko supporters are going to hear a lot about the polls that show him in second, including the TPP’s internal polling (which is not included in our poll of polls), and very little about or next to nothing about the polls showing him far behind. Their internal polling as well as a poll by Credere Media show Ko behind Lai by a very low margin.
TPP supporters also like to tout street polling, which is not included in the poll of polls, which generally show Ko's support far higher than most polls. For example, ‘Streets have Parties’ (街頭有派對) spoke to 1900 people in person Dec. 9-26 and found Lai was still in the lead with 37.5%, Ko a close second with 35.9%, and Hou last with only 26.6%.
There are enough polls out there that show Ko in second, and some with him even having a chance to win, that there is no reason for his supporters to strategically vote for someone else. There is also an added incentive to vote for Ko, which is that even if he comes in second, it sends a strong message that there is a new force on the scene that wants a change from the two bigger parties, which is a victory of sorts. That is not true for supporters of the two established parties.
Polling problems
There is another thing to consider: what if the polls showing Ko is actually in second are more accurate? This possibility can not be ruled out, and I am not the only one raising this possibility.
Traditional polling used landlines, but these days few young people have them, so their numbers are underrepresented in the overall sample. To correct this, pollsters weigh the few younger respondents higher, but this potentially distorts the picture.
Cell phone polls are growing more common, but this presents the opposite problem as older voters are underrepresented. Many outfits are now conducting half landline, half cell phone polling in hopes of balancing the two out.
Yet other pollsters use various digital methods to reach out to voters. RW News uses a mix of SMS text messages, emails, and other digital methods to get their data. During the recent local elections, their results were not too bad, if not great.
Street polling is not included in the poll of polls because it is hard to be certain whether they are genuinely trying to get an accurate sample or not. However, it is possible they managed to do so.
Cell phone polling, digital, and street polling show Ko’s support higher, whereas more traditional pollsters show Ko trailing badly. It will be very interesting to see which types of pollsters did better after the election and to adjust the poll of polls accordingly if necessary.
Ko could win
If it is unlikely that Ko’s supporters will abandon him at the ballot box, could the supporters of other candidates choose to vote strategically? Lai’s supporters will not, as he is ahead in all the polls.
Hou’s supporters probably will not either. His supporters are older and consume more traditional media, which tends to report more on traditional polling, which has him running in second, giving his supporters no reason to throw their support behind third-placed Ko.
It is theoretically possible that Ko could win if the polls that are showing him a close second are accurate, and he wins over more of the undecided voters. I estimate roughly five to eight percent of the population that is in the undecided column will turn out to vote, with the rest staying home.
Ko would have to win a solid majority of those to overtake Lai. That is a steep hill to climb, but not impossible. Unlike Hou or Lai, he does not have as much political baggage, which may make him appealing.
Regardless, the actual election results will teach us a lot about polling in Taiwan. We will see if Ko and his supporters are right they are being undercounted. Or not.
