.jpg){kind=link}
The answer: another signal, not a completed miracle
Z.Media’s October 22 Taipei mayoral release reports Puma Shen at 42.8% and Chiang Wan-an at 39.0%. It adds preference evidence from a different research operator to discussion of the Ocean Current. It remains a survey result, not a victory already secured for November 28.[1][6]
Senge’s next test is whether the signal survives reasonable designs, becomes participation and faces scrutiny of governing proposals. Declaring the outcome early and rejecting every poll both skip the analytical work.
Understand the design before comparing pollsters
The publisher says Z.Media funded the survey, Shan Shui conducted it on October 18–19, and 1,074 registered Taipei residents aged 20 or older responded. It used a residential-telephone frame and demographic weighting, with a published single-proportion margin of about ±2.99 percentage points. Disclosure helps identify the evidence; it does not remove every possible bias.[6]
Subtracting this margin from another pollster’s different field period cannot establish how many people switched in two days. Frames, candidate lists, question order, response and weighting can affect comparison. Demographic adjustment addresses included characteristics, not every unobserved nonresponse difference. Recognizing that limitation also does not prove the result wrong.
Three numbers answer different questions
| Observation | This wave / source | Useful question | Unsupported extension |
|---|---|---|---|
| Overall preference | Shen42.8%, Chiang39.0% [1][6] | Preferences under this design | Neither certified votes nor winning odds |
| Certain-to-vote subgroup | Shen51.2%, Chiang38.6% [2][3] | Preferences within a stated-intention subgroup | Not completed votes or a validated turnout model |
| Expected winner | Chiang46.4%, Shen36.9% [1] | Whom respondents expect to win | Not their own vote choice |
| Recalled past vote | Reports cross-tabulate recalled2022 choices [3][4] | Recall and current response in this sample | Not longitudinal tracking of the entire electorate |
Preference and expectations can point opposite ways
TVBS reports Chiang ahead on expected victory but Shen ahead on preference. Both can be true: someone may want one candidate and expect the other to win. Combining expectation, personal choice and probability would manufacture a contradiction.[1]
Expectations could reflect impressions of organization, incumbency or surrounding discussion, but this evidence cannot identify the cause. Nor can it establish that lower expectations necessarily mobilize supporters. The useful follow-up is whether comparable questions show preferences and expectations converging, and how those patterns relate to participation.
Certain intention is not a recorded vote; recall has limits
Mirror Daily and Mirror Media report the certain-to-vote subgroup. That can inform questions about mobilization, but selection rests on stated intention rather than final voting records. Without full subgroup bases, weights and model documentation, it cannot be declared more accurate than the full sample.[2][3]
Likewise, the published ±2.99 describes an overall single proportion, not a significance threshold for the candidate difference or every subgroup. The variance of a difference depends on the relationship between estimates; subgroups also generally have smaller effective samples. This article makes no unsupported significance claim.
Nextapple and other reports interpret preferences through recalled earlier votes. A current recall cross-tab does not follow every voter from four years ago. Memory, selective response and sample composition can intervene. Multiplying a subgroup percentage by a historical vote total would not establish an exact number of transferred ballots.[4]
Agreement and differences among sources
TVBS, Mirror Daily, Mirror Media and Nextapple agree on the central figures while drawing mainly on one publisher’s wave. They help compare reporting, not supply four independent measurements. Methods should be traced to the publisher; unread image tables cannot be described as fully checked.
CNA’s October 20 interview supplies different evidence: responding to the earlier Shinmin poll, Shen emphasized participation and policy resonance. This is a candidate’s position, not independent research proving his prospects or policy effects.[5]
Plausible alternatives include actual preference movement, design differences and differences in response or participation. Available material cannot allocate precise shares to those explanations. Balance means applying the same evidentiary standards to favorable and unfavorable findings, not demanding identical numbers for both candidates.
Conclusion: turn attention toward inspectable choices
Another wave gives further reason to examine the Ocean Current, without removing methodological and turnout tests. Watch whether similar designs reproduce results, whether pollsters show a consistent direction and how candidates explain costs, responsibilities and implementation timetables. These are questions for further checking, not effects already measured by Senge.
Attention and discussion may begin participation. Analysis must preserve the distance between observation and conclusion. Useful election commentary tells readers what the evidence establishes and what still needs an answer.