A back-of-the-envelope guesstimate

Where does the presidential race stand now? Eyeballing the most recent polls, it looks to me like it is roughly Tsai 44, Chu 22, and Soong 11. Conveniently, that is a neat-looking 4-2-1 pattern. Ok, but what does that mean for actual votes? Is Tsai likely to break 50% in the actual votes? If so, by how much?

Each election is a new soap opera, so there is a limit to how much we can learn by looking at old results. Still, it would be nice to see what similar polling results produced in the final vote tallies. I’m looking for races, preferably with lots of poll taken close to election day, in which one side was clearly ahead For polls, I will look at the last few polls listed in my list of survey results on the sidebar of the front page of this blog. As above, I’m going to eyeball the results and blurt out a rough average.

race candidates Polls Votes Increase
2014 Taipei Lien (K) 30 41 11
Ko (I) 45 57 12
.
2014 New Taipei Chu (K) 49 50 1
Yu (D) 28 48 20
.
2014 Taoyuan Wu (K) 49 48 -1
Cheng (D) 28 51 23
.
2014 Taichung Hu (K) 28 43 15
Lin (D) 44 57 13
.
2014 Tainan Huang (K) 16 27 11
Lai (D) 66 73 7
.
2014 Kaohsiung Yang (K) 16 31 15
Chen (D) 58 68 10
.
2014 Keelung Hsieh (K) 15 27 12
Lin (D) 38 53 15
Huang (I) 11 16 5
.
2010 Tainan Kuo (K) 25 40 15
Lai (D) 48 60 12
.
2010 Kaohsiung Huang (K) 13 21 8
Chen (D) 45 53 8
Yang (I) 25 27 2
.
2008 president Ma (K) 52 58 6
Hsieh (D) 30 42 12

A few patterns stick out. First, New Taipei and Taoyuan were completely wrong in 2014. They weren’t heavily surveyed, but all the survey data we had showed the two KMT candidates way ahead. [Edit: This was also the case in the Taiwan Election and Democratization surveys for those two cities. Oops. I am working with the TEDS data today, and I was wrong. The TEDS surveys, which were done in early November 2014, show Chu with only a 8% lead in New Taipei and a dead tie in Taoyuan. If I had seen these results before the election, I wouldn’t have been so shocked on election night. However, none of the media polls showed these to be close races.] Sometimes things go in a completely unexpected direction.

Second, what usually happens is that the main candidates all increase by about the same amount, especially in two candidate races. It is rare that one candidate increases by twice as much as the other main candidate.

Third, in the two three candidate races, the independent candidates got a much smaller bounce than the major party candidates.

Fourth, trailing candidates tend to make up a little ground on the leaders, but not too much.

Intuitively, there are compelling reasons for these patterns. On the one hand, a disproportionate amount of undecided voters will decide to stay at home. If they don’t vote, everyone else’s vote percentage increases proportionally. Trailing candidates often have a somewhat larger pool of undecided voters they can appeal to. If they are unknown or a bit personally unpopular, it may take some wooing to convince party sympathizers to vote for them. On the other hand, many other undecided voters are actually neutral. When it comes time to vote, it would be unreasonable to expect them all to swing the same way. Some will go one way at the last second, and others will opt for the other candidate.

What does this all imply for this year’s election? Keep in mind that we aren’t at the end of the campaign yet. I expect that Soong’s support will further erode between now and January. I don’t have clear expectations for the other two candidates, but I’d be surprised if the polls don’t move at all over the next three months. Still, if this were the end of the campaign, I’d expect Soong’s 11% to stay about the same and for Chu to take slightly more of the remaining 23% than Tsai. Let’s say that Chu would get 13 and Tsai 10. That yields Tsai 54, Chu 35, and Soong 11.

As I said above, I expect that Soong will continue to fade, and polls show that his supporters will go about half and half to the other two candidates. Thus, my best guess right now is that the election will be roughly Tsai 57, Chu 38, Soong 5.

Keep in mind that these calculations were all done very roughly, and so this is a very imprecise guess. The big point is that it isn’t that unreasonable to expect Tsai’s current 20% (or so) lead in the polls to translate into a 20% (or so) victory in the final vote tally. Of course, every once in a while something really strange happens, so don’t write this in stone.

4 Responses to “A back-of-the-envelope guesstimate”

  1. Mike Says:

    More of a technical question, but do you know roughly know what sort of methods are currently “trendy” among polling agencies when converting support rates (支持度) to voting projections (得票率預估)? Do they look at intention to vote, past voting preference? Or do they put other variables into the modelling?

    • frozengarlic Says:

      I think all of them throw away all the people who say they don’t plan to vote. The big question is what to do with people who say they will vote but haven’t yet made up their minds. Some agencies simply throw these away and recalculate with the denominator as only those who express a preference. However, most will try to run some sort of statistical model to predict how the undecideds will vote. Typically this uses their answers on other questions, such as demographics, party ID, national identity, preference on unification/independence, or other issue preferences. In contrast, the political map approach asks them where they live and uses the past voting history of their neighborhood to assign the candidates fractions of their votes.

      Personally, I usually ignore predictions unless I know how they are done. Predictions require a lot of assumptions, and analysts (who are human and naturally want to see good news for their favorite candidate) are prone to produce results that please them. I’d rather see the raw data.

  2. omega Says:

    what went wrong with the polling in New Taipei and Taoyuan last year ?

    • frozengarlic Says:

      I don’t know. In a way, it is even more puzzling that the TEDS polls showed close races in early November, Otherwise, we could speculate that the races tightened up in the last ten days during the polling blackout period.

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 )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s


%d bloggers like this: