In August, SPI tested which policy messages deserved to be said first. In September we went back to the same voters to see whether they had started to choose. The result: the undecided pool shrank only slightly, and women and working-age voters remain the group most worth understanding. For a candidate, the question is no longer just "what do we say" — it is "to whom, and how do we make the promise part of their daily life".
In August's Same issue, different frame: what 1,003 responses reveal, we compared twelve policy messages and identified the directions most worth communicating first — night-time childcare, medical care for young children, denture subsidies for older adults, and youth employment and start-up support.
In September we did not ask the same questions again. We returned to the same voters and watched whether they had begun to make a choice. Together the two studies form one decision chain: August answered "what to say first"; September goes on to answer "whom to say it to first, what problem to solve, and where to reach them".
Among the 404 six-city voters who completed both the August and the September waves, the share who had not yet decided fell from 33.4% to 31.7% — a drop of only 1.7 percentage points.
This is not a rapid return to party lines. Most undecided voters are still waiting for a reason good enough to back a candidate: they have not clearly walked away, and they have not drifted naturally toward anyone. For campaign decision-makers, time on its own will not do the persuading. A promise has to show voters how choosing this candidate changes their own lives.
New Taipei illustrates the point. Among the same New Taipei voters, the undecided share edged up from 28.9% to 29.8%. A heating-up race did not shrink the undecided pool by itself.
The map uses the full September wave (682 weighted respondents), so its city figures differ slightly from the 404-voter same-person changes above. All six cities are shown: Taichung 41.3%, Tainan 31.5%, New Taipei 30.2%, Kaohsiung 29.6%, Taoyuan 26.1%* and Taipei 12.1%*. *Low-sample, directional reads (19 and 21 undecided respondents; roughly ±12 and ±8 points) — read them with their ranges, not as rankings. The interactive map is in Traditional Chinese; each city's range and sample are shown on click.
Among the same women, the undecided share fell from 42.6% to 38.2%; among men it moved from 24.0% to 25.0%.
Women did move — but in September close to four in ten still had not decided. That makes them the clearest and largest persuasion opportunity available today. What they need is not a generic "we care about families" slogan but concrete commitments that answer the bills, the job, housing and caring responsibilities.
Working-age voters show a similar picture:
Both groups are at the most expensive stage of life: work, starting a family, raising children, buying a home and caring for parents often arrive at the same time. The policy directions that tested strongest in August are worth continuing precisely because they answer these everyday pressures.
In September, undecided voters most often named four issues:
| Priority issue | Undecided voters naming it | All six-city voters naming it |
|---|---|---|
| Cost of living / inflation | 51.5% | 46.2% |
| Wages / jobs | 32.3% | 27.2% |
| Housing | 24.7% | 21.2% |
| Public safety / fraud | 18.8% | 17.1% |
All four run higher among undecided voters than across the six cities as a whole. The common signal is clear: these voters are not short of political information; they have not yet heard an answer good enough to change their choice.
A policy promise therefore has to make three things instantly understandable: who gets helped, how much pressure it takes off each month, and when the change will be felt. "Improve the economy" or "value young people" win easy agreement but rarely separate one candidate from another.
Among undecided voters, 53.6% said that if a candidate's position on nuclear power were the opposite of their own, they might or would definitely reconsider their vote; a further 20.9% did not know where the candidate stands.
The two numbers point to different risks. Some voters already hold a position, and a candidate who avoids the topic may still pay for it late in the race. Others do not even know what the candidate proposes — the position has not been clearly received.
What needs testing is not merely "for or against nuclear" but how the candidate explains the trade-off between electricity prices, supply stability and safety. The more complex the position, the simpler the explanation has to be.
Undecided voters' main sources of election news are:
Respondents could choose more than one. What the figures show is not a single "best channel" but a path along which a message travels: news sites and television establish credibility first, LINE carries the message into networks of people who know each other, and Facebook and YouTube add the person, the story and the explanation.
If the whole budget goes into short video, a message may have volume without producing understanding. For a professional team, the heart of media planning is whether one promise stays consistent across channels — not a different story for each platform.
August's message test found the policy directions worth investing in; September's tracking shows that many voters have still not decided, and that the people most worth understanding first are women, voters aged 20–59 and those under everyday financial pressure.
This is the value of continuous tracking. A single survey describes a moment; repeated measurement of the same people tells decision-makers who is actually moving, which questions remain unanswered, and where the next resource should go.
If you are a candidate's team, a campaign director or a campaign decision-maker and want to confirm the undecided audience, the message gaps and the priority channels in your own constituency, you are welcome to book a private 45-minute briefing: sales@sharepartyinsight.com.
September did not re-test August's twelve messages, so this article does not attribute this wave's changes to any individual policy message.
Earlier articles in this programme: 35.6% undecided: Taiwan's first six-city election-strategy wave · Same issue, different frame · Challenging an incumbent: continue, improve or confront? · Undecided voters are the largest pool; the digital contest has barely begun