.jpg){kind=link}
The answer: better performance and lower preference can coexist
After Taipei’s debate, TVBS reports Chiang Wan-an46% and Puma Shen42% on vote preference, while Shen leads perceived performance. Performance, liking an argument and choosing a mayor are different questions.[1][6]
For the Ocean Current, the next test is whether attention reaches beyond existing supporters and survives scrutiny of governing choices. A survey conducted after a debate is not automatically a measurement of change caused by it. Sequence begins the inquiry; it does not establish causation.
First test: identify the population behind each number
TVBS interviewed1,747 Taipei residents aged20 or older between2pm and6pm on October31, using interviewer-administered residential telephone random-digit sampling. TVBS funded the survey.[6]
Original tables base preference on1,666 respondents expressing voting intention; performance evaluation uses all respondents. Subtracting percentages across those bases cannot identify lost convertible votes. Stated intention is not a voting record.
No-opinion and undecided categories differ too. Someone without a performance judgment may have a party preference; someone impressed may remain undecided. Without the full joint distribution, aggregate percentages cannot tell each respondent’s story.
Three tests and the evidence needed
| Test | What this wave measures | What it cannot establish | Useful next evidence |
|---|---|---|---|
| Population and question | Intenders’ preference; all respondents’ performance judgment [6] | A percentage gap as lost votes | Question wording, bases and joint responses |
| Viewing and choice | Different preferences across viewing groups [3][4] | How many switched because they watched | Prior preferences, panel tracking or stronger identification |
| Timing and effects | An after-debate snapshot [1][2] | Every multiweek change caused by the debate | Comparable nearby waves and checks of other events |
Second test: viewing is self-selected
NOWnews and SETN emphasize differences between viewers and nonviewers. Those groups may already differ in interest, preference, schedules and media exposure. A post-event difference does not identify when it arose.[3][4]
Possible explanations include Shen supporters being more likely to watch, some viewers being persuaded, or both. Without prior preferences, none can be declared the sole explanation. That limitation does not prove the debate had no effect; it prevents separating persuasion, mobilization and selection precisely.
Evidence of reaching new voters would identify prior nonsupporters, why they changed and whether change persisted. This is Senge’s proposed research question, not tracking already conducted by this site.
Third test: same-pollster comparison still has limits
Business Today and Mirror Media describe a narrower gap than early October. A shared publisher helps comparability, but intervening events and sample composition remain. A survey does not automatically control them like an experiment.[2][5]
The original tables also juxtapose expected debate performance before the event with evaluation afterward. Expectation and retrospective judgment are different constructs; subtraction cannot establish individual conversion. We estimate neither persuasion nor the publisher’s turnout model.
The overall single-proportion sampling margin is not a universal significance threshold for candidate differences or subgroups. Telephone coverage, nonresponse and weighting still matter; a seemingly high response rate cannot remove every unobserved difference.
Agreement across five reports and the analytical boundary
TVBS, Business Today, NOWnews, SETN and Mirror Media share one wave. Their matching figures are not five independent validations of a debate effect. Different headline choices highlight preference, performance or viewing groups; original tables clarify the question being answered.
Mirror Media’s displayed time precedes the stated interview end. We do not use it to infer publication sequence or a separate wave, and rely on TVBS’s original for figures. Turning-point and mobilization language is interpretation, not causal identification.[5][6]
Balance applies the same standard to favorable and unfavorable signals. Shen’s performance advantage does not guarantee victory; Chiang’s preference lead does not remove policy accountability. Renewal, services and fiscal promises still require checks of responsibility, cost, timing and observable outcomes.
Conclusion: make attention an inspectable public choice
Voters can approve performance, reserve their choice and remain uncertain about the outcome. The Ocean Current needs durable trust beyond attention; an incumbent must turn experience into specific answers.
Before November28, the useful test is who makes commitments inspectable, rather than who collects victory headlines. Clear populations, selection mechanisms and timing help commentary improve judgment without writing an unfinished election’s ending.