Taiwan’s Six-City Election Tracker: 35.6% Undecided—and Which Bases Are Actually Firm?

SPI Political Knowledge Lake | First wave across Taiwan’s six largest municipalities

Figure note: The combined results use 834 eligible respondents. SPI first weighted each municipality to its voter profile, then combined the six using each municipality’s electorate. The result is not a simple average of six percentages. On this basis, 35.6% were undecided; the top-two gaps in New Taipei and Kaohsiung were 0.9 and 0.4 percentage points; and the share of each candidate’s supporters classified as firm ranged from 51.1% to 93.3%. The next wave will track how these figures move. Graphic by SPI.

Key findings

Across the weighted six-city sample of 834 eligible respondents, 35.6% had not chosen a candidate and 10.7% explicitly selected “will not vote.” The top two were separated by only 0.9 percentage points in New Taipei and 0.4 points in Kaohsiung. Among the people already supporting a candidate, the share still needing to be consolidated ranged from 6.7% to 48.9%. Women were 41.7% undecided, while only 35.1% of respondents aged 20–39 rated themselves highly likely to vote. The strategic question is therefore not simply who leads today, but who can move undecided voters and turn expressed—but not yet locked-in—support into turnout.

By Tatt Chen, Founder and CEO, ShareParty Insight (SPI)

This was SPI’s first six-city wave, conducted from 31 July to 3 August 2026. After duplicate responses were removed, the survey contained 1,036 valid responses. The voting results in this article use only the 834 respondents eligible to vote in local elections in Taiwan’s six largest municipalities. This is not a forecast of November’s winners. It is an early read on which candidates have stable support and where votes may still move.

Three questions that matter

1. Whose current support is most firmly consolidated? Among Chen Ting-fei’s (陳亭妃) supporters in Tainan, 93.3% were both highly likely to vote and highly certain of their candidate choice—the highest share in the study. Ho Hsin-chun (何欣純) in Taichung followed at 80.8%, and Lai Jui-lung (賴瑞隆) in Kaohsiung at 76.7%. Chang San-cheng (張善政) registered a large 31.7% early lead signal in Taoyuan, but only 51.6% of his current supporters were classified as firm; almost half could still move.
2. Which small groups should be accumulated over time rather than overinterpreted now? This wave included only 41 respondents who were not yet eligible to vote in 2022 and 43 Taiwan People’s Party voters. SPI uses both only for internal directional observation, not headlines or principal conclusions. Later waves can be assessed once enough cases accumulate.
3. Who may move, and what should be tested next? Women, 20–39-year-olds, each candidate’s not-yet-firm supporters, and voters who have not chosen a candidate warrant the closest tracking. The next wave should directly test whether messages can lift issue agreement, candidate certainty and turnout intention.

• 35.6% had not chosen a candidate.
• 10.7% explicitly said they would not vote.
• Only 49.8% rated themselves highly likely to vote (7–10 on a 0–10 scale).
• Using all 834 eligible six-city respondents as the denominator and grouping the questionnaire’s named candidates by party, KMT candidates totalled 22.8% and DPP candidates 18.2%. The undecided share was larger than either party’s total. This is not party identification; it is the weighted response to “who would you be most willing to vote for now?”

The implication is straightforward: the first-place name alone is an incomplete picture. Persuading undecided voters—and converting soft declared support into actual turnout—may matter more than an early ranking.

Three conditions across the six cities: leads, close contests and undecided voters

Figure 1: Each city percentage uses respondents from that city as its denominator. Candidate options reflect the questionnaire at the time of fieldwork. These are August early-support signals, not forecasts of November results.

Taipei and Taoyuan comprise the first condition: relatively large early gaps. New Taipei and Kaohsiung comprise the second: the top two are effectively tied. The third condition exists in every city: at least 32% were undecided, reaching 42.5% in Tainan. An August lead is therefore only a “lead signal,” not a November result.

The 0.9-point gap in New Taipei and 0.4-point gap in Kaohsiung should be read as “currently almost even,” not as evidence that either side has won. Taichung’s 5.9-point and Tainan’s 8.9-point gaps likewise remain early differences to track in the next wave.

“Explicit non-vote” means respondents who directly selected “will not vote.” It does not include everyone who rated turnout likelihood from 0 to 6. Kaohsiung (13.4%) and New Taipei (13.0%) were highest. In cities separated by 0.4 and 0.9 points, getting supporters to turn out may matter more than the current rank order.

Whose support is firm—and whose still needs consolidating?

SPI divided respondents who selected a candidate into two groups. “Firm” supporters rated both turnout likelihood and certainty of candidate choice at 7–10. “Soft” supporters rated at least one of these at 0–6. The latter group has insufficient turnout intention, insufficient certainty, or both; it does not mean every person in it will abstain.

The table ranks candidates by the share of current supporters who were both highly likely to vote and highly certain of their choice. “Needs consolidating” is the complement: people who named a candidate but rated turnout likelihood or candidate certainty at 0–6. They may not vote, may change choice, or both. The measure must not be interpreted as a definite non-vote rate.

The sharpest contrast is between Taoyuan and Tainan:

• Chang San-cheng stood at 31.7% in Taoyuan, but only 51.6% of his supporters were firm. As shares of the full city sample, firm support was 16.4% and soft support 15.4%—almost equal.
• Chen Ting-fei stood at 24.9% in Tainan, with 93.3% of her supporters firm. As shares of the full city sample, firm support was 23.3% and soft support only 1.7%.
• In New Taipei, the candidates’ total support was close, but Su Chiao-hui’s firm share of 72.3% was slightly above Lee Shu-chuan’s 67.4%.
• In Kaohsiung, Lai Jui-lung’s firm share of 76.7% exceeded Ko Chih-en’s 60.5%, although their total support differed by only 0.4 points.

A lead is not necessarily secure, and trailing does not necessarily rule out convergence. Tracking only topline support in the next wave would miss much of the movement underneath.

Small groups: no flow claims yet

This wave included only 41 respondents who had not been eligible to vote in 2022 and 43 Taiwan People’s Party voters. Even though both reached this article’s minimum disclosure threshold for subgroups, they are still too small to support a headline, a principal finding or a one-decimal-point flow estimate. SPI therefore does not publish party-flow percentages for either group in this release. We will reassess after later waves accumulate enough observations.

Who may decide the race? Women, younger voters and unlocked support

Figure 2: These are comparisons within the current sample; they do not represent the entire electorate.

Among women, 41.7% had not decided, compared with 29.1% of men. Women also more often selected “will not vote”—12.7% versus 8.5% among men. The question is therefore not only whom women support, but whether they have entered the contest at all.

The age gap is more direct. Only 35.1% of respondents aged 20–39 rated their turnout likelihood at 7–10, compared with 57.2% among those aged 40–59 and 56.9% among those aged 60 or older. Younger respondents were also 35.7% undecided and 14.1% explicit non-voters. For every party, the first question about younger voters may be whether they turn out—not which side they choose.

Prior party leaning and current candidate choice: about two-thirds of DPP and KMT identifiers stayed with their party

SPI Knowledge Lake retained all 834 eligible six-city voters in the candidate-choice analysis, then linked each anonymous Panelist ID to that person’s latest usable party-leaning response. People with no stated affiliation, a refusal or no historical answer remain separate groups; SPI does not infer a party for them.

The table shows only the two major-party candidates, undecided and explicit non-vote. Other candidates, minor-party candidates and refusals are omitted, so rows do not necessarily sum to 100%. Other minor-party samples are not yet large enough for stable interpretation and are not reported separately; SPI will publish them only after enough observations accumulate.

Roughly two-thirds of prior DPP and KMT identifiers currently chose their party’s candidate, indicating a real but not fully locked base. The larger opening was among people with no party leaning: 45.3% were undecided, 18.5% explicitly would not vote, and only 39.5% rated turnout likelihood at 7–10. The 43 Taiwan People’s Party voters are not used for a public flow conclusion in this table.

Five signals to track next

1. Do New Taipei and Kaohsiung remain effectively tied? Their current gaps are only 0.9 and 0.4 points.
2. Can Taoyuan’s lead become stable support? Firm and soft supporters are almost evenly split within Chang San-cheng’s current support.
3. Where does Tainan’s 42.5% undecided share move? It leaves considerable room for the early order to change.
4. Do women and younger voters enter the contest? These groups stand out for undecided status or lower turnout intention.
5. Do small groups become large enough for defensible public interpretation? First-time voters and Taiwan People’s Party voters remain small in this wave.

What comes next

SPI plans one September follow-up: place the next wave beside this one and track changes in undecided voters, firm support, turnout intention and issue priorities across the six municipalities.

About SPI and the author

ShareParty Insight (SPI), founded and led by Tatt Chen, is a data-advisory company combining survey research with consented mobile-use behaviour. All data collection is consented and authorised. Analyses are released only as aggregate results; personal data are not published.

Tatt has more than 20 years of research, data and corporate experience. He previously served as Asia Senior Vice President at Penn, Schoen & Berland (PSB), a WPP research consultancy, leading its Hong Kong office and Asia business.

Further reading: Same issue, different frame—what 12 policy messages reveal about campaign priorities

SPI Taiwan Political Intelligence

For a briefing on what these signals mean for your Taiwan decision, contact sales@sharepartyinsight.com.

This article was drafted with assistance from AI, but edited, verified and published by ShareParty Insight.