Evaluating TVBS

I posted some TVBS polls on the Feb By-elections a couple of posts below and a friend suggested I take a look at TVBS polls (poll center) from the Dec 5 county/city elections to get a sense of how accurate the TVBS polls are. Each number means the percent by which the KMT is ahead of the DPP in the poll (ex: +5 = KMT ahead by 5%). Outcomes are from CEC data (PDF). All the polls were of likely voters. The elections are all county elections.

I-lan
KMT vs DPP (four polls from Feb, Oct, Nov & Nov): +12, +7, +5, +6
Actual outcome: DPP +8.5 (KMT -8.5)

Hsinchu
The Hsinchu race featured three candidates. Three polls from Oct, Nov, and Nov:
KMT over Ind/DPP: +3/+16, -4/+12, +3/+12
Actual Outcome: +8 over both IND and DPP

Chiayi:
Two polls from Oct and Nov
KMT over DPP: -5, -8
Actual Outcome: DPP +15 (KMT -15)

Yunlin:
Two polls from Oct and Nov
KMT over DPP: -33, -35
Actual Outcome: DPP +31 (KMT -31)

Nantou
Two polls from Oct and Nov
KMT over DPP: +22, +31
Actual Outcome: KMT +11

Hualien:
Two polls from Oct and Nov
KMT over IND: -9, -25
Actual Outcome: IND +49 (KMT -49)

Taoyuan
One poll from Nov
KMT over DPP: +38
Actual Outcome: KMT +6.6

Miaoli
One poll from Oct
KMT over DPP: +53
Actual Outcome: KMT +30

Taidong:
One poll from Nov:
KMT over DPP: +29
Actual Outcome: KMT +5

Changhua
One poll from Oct:
KMT over DPP: +22
Actual Outcome: KMT +11

Some consistent patterns: except where the DPP has a blow-out lead, TVBS always underestimates the size of the DPP vote, sometimes by very large margins. Another way to look at it is that its findings are strongly biased for the KMT, which is not to say the pollsters are biased, but rather than its findings appear to be. Perhaps KMT voters are more likely to express a willingness to vote for the KMT to a Chinese-owned TV station that is ardently pro-KMT, or for some reason the undecideds predominantly break DPP. Another consistent pattern is that in almost every poll that is not a massive blowout, undecideds constitute 25-35% of the sample. In sum, unless the KMT candidate has a lead of at least 10 points, s/he should probably be considered within reach.

Thus looking at the early polls for the legislative by-elections in Feb, Hualien (KMT+19) should be regarded as out of reach for the DPP (but look for that gap to close), Taoyuan (KMT +3 over DPP and +21/28 over independents) is in play and might even be DPP win, while Hsinchu (KMT -10) looks like a pretty safe DPP bet, and Chiayi (KMT -31) is going to be a DPP blowout victory.

A closer look at the Taoyuan election. The election is being held in Taoyuan 2. That district consists of four townships: Tayuan, Yangmei, Hsinwu, and Guanyin. Rounding down, in the county chief election, they went DPP +8, KMT +12, DPP +13, and DPP +15. The total vote count across those was roughly 70,000 votes for the DPP and 67,000 for the KMT. But all politics are local, and legislators are different than township chiefs, so don't set too much store by that. There are two independents in the race, poaching votes.

Further point: A longtime analyst and observer here told me that in the educated demographic in Hsinchu, the DPP's candidate, highly educated, blows away the KMT candidate.
_________
DAILY LINKS:

• "Isolated" Iran, Russia, China in key gas deal.
• Dixteel says KMT isn't really right-wing
• Ma to transit through US.
• Epoch Times commentary rips Friedman's Taiwan piece. Really, it's too easy.

_______________________
[Taiwan] Don't miss the comments below! And check out my blog and its sidebars for events, links to previous posts and picture posts, and scores of links to other Taiwan blogs and forums!