Will Chiang Wan-an win the next Taipei Mayor election?

TL;DR

Chiang Wan-an’s probability of winning the next Taipei mayoral election is 46.7%; the probability of No is 53.3%. He is a slight underdog in a close contest, not a secure incumbent. Treat conflicting polls and turnout uncertainty as the central facts, rather than taking any single survey as the answer.

Context

As of October 31, the race is approaching its final four weeks. Voting is scheduled for November 28, 2026, and the question concerns who becomes Taipei mayor as a result of that election. An interim appointment does not count. The June 30, 2027 fallback is a deadline for an unresolved outcome, not an extension of the campaign. (Official election calendar)

The evidence describes a competitive race between Chiang and Shen Po-yang. Some surveys put Chiang ahead; others put Shen ahead. The newest survey preserves a path to Chiang’s reelection, while the longer trend and supporter-engagement evidence explain why incumbency alone does not make him the favorite.

Evidence

The historical starting point favors incumbents, but the reference class is small. Elected Taipei incumbents won three of four reelection attempts from 1998 through 2018: Chen Shui-bian lost, while Ma Ying-jeou, Hau Lung-bin and Ko Wen-je won. That record supports an initial incumbency advantage. Four different contests do not establish a rule that overrides current polling. (Taipei government history)

Chiang’s own electoral base also needs context. He won the November 26, 2022 election with 575,590 votes, or 42.29% of 1,360,951 valid votes. That was a plurality in a three-way contest, not a standing majority that transfers automatically to a largely two-candidate race. The result demonstrates electoral strength, but it leaves the allocation of former third-candidate voters open. (CEC-sourced election record)

Past polling does not provide a reliable automatic correction in Chiang’s favor. TVBS’s November 2022 closing survey reported Chiang at 36%, Chen at 25%, and Huang at 24%. Chiang ultimately received 42.29%, against Chen’s 31.93%. Both leading candidates gained relative to those raw poll shares; the survey’s Chiang–Chen lead was slightly larger than the final gap. The three-way contest complicates comparison, but it does not support assuming that polls systematically hide an incumbent advantage. (Closing poll and comparison)

The strongest current evidence is the disagreement across surveys. TVBS’s October 31 poll reports Chiang at 46%, Shen at 42%, and 12% undecided. It used 1,747 landline interviews, with the voting-intention table covering 1,666 willing voters. This is direct evidence that Chiang remains competitive. It is not evidence that his lead is secure. (TVBS report)

The full TVBS trend available here points toward a much tighter race. Chiang’s reported lead was 28 percentage points in May. In the October 3–7 survey, he led 49% to 41%, an eight-point margin. On October 31, the margin was four points. Comparing one organization over time reduces some differences in method, but these are separate samples, not repeated interviews with the same voters. The trend establishes narrowing; it does not establish another month of decline. (TVBS historical comparisons)

UDN points in the opposite direction from the newest TVBS result. Its October 28 release, based on fieldwork on October 20–22 and 25, reports Chiang at 41% and Shen at 45%. It interviewed 1,115 respondents using landline and mobile phones. ETtoday’s October 27 release reports Chiang at 48.6% and Shen at 41.1%, based on 1,462 respondents surveyed October 21–25 through an invited online membership panel. These surveys finished on the same date yet named different leaders. Their disagreement is therefore not a simple story of one candidate rebounding after the other poll ended. (UDN; ETtoday)

Another survey, Zmedia/Shanshui, reports Chiang at 39.0% and Shen at 42.8%, based on 1,074 landline interviews conducted October 18–19 and released October 22. Its original release was not verified directly, so it deserves less confidence than the primary reports. Even so, its direction reinforces the point that the favorable TVBS and ETtoday readings are not the whole picture.

Beacon provides the most adverse result for Chiang. Its October 14–16 survey of 1,075 respondents—753 landline and 322 mobile interviews—reports Chiang at 33.3% and Shen at 50.3%. That raw result is genuine. But the pollster’s turnout and past-vote scenarios imply a range from Chiang leading by 0.7 percentage points to Shen leading by six. Those are assumption-dependent projections, not confidence intervals. The size of the change between the raw responses and projected electorate makes turnout composition a central uncertainty. (Beacon report; projection coverage)

Several weaker signals help explain the slight disadvantage. UDN reports election engagement among 91% of Shen supporters, compared with 77% of Chiang supporters. Its recalled-vote cross-tab puts former Huang Shan-shan voters at 55% Shen and 38% Chiang. TVBS reports a similar split, 53% to 39%. These subgroups are vulnerable to recall and selection bias, but they challenge the assumption that former third-party voters naturally consolidate behind Chiang. (UDN; TVBS)

The debate does not establish a Chiang recovery. TVBS respondents preferred Shen’s performance, 47% to 40%; neutral respondents preferred Shen, 44% to 24%. Those findings indicate a vulnerability, not a measured causal vote swing. Viewers chose whether to watch, and the survey cannot isolate what the debate changed. (TVBS debate findings)

Campaign events cut both ways. October 31 reporting describes continuing objections to police conduct during a gay-bar search, Chiang’s renewed apology, and protests during his Pride appearance. That creates a plausible mobilization issue, but demonstrators are not a representative electorate and the controversy may already be reflected in polls. Meanwhile, senior KMT figures and TPP chair Huang Kuo-chang appeared at Chiang’s October 26 campaign launch. The claimed crowd exceeded 50,000; that was an event claim, not a verified count of Taipei voters. The evidence shows an active organization, not organizational collapse. (CNA; Public Television)

What's non-obvious

The key distinction is between party alignment and voter transfer. UDN reports current TPP identifiers favoring Chiang 75% to 16%, while recalled Huang voters favor Shen 55% to 38%. Those groups are not interchangeable. Reading party-level cooperation as proof that Chiang inherits the previous third-party electorate misses a potentially decisive source of opposition support. (UDN)

A second distinction is between recency and representativeness. The newest poll deserves attention, but a newer landline survey does not automatically settle disagreement with a dual-phone survey or an online panel. Likewise, Beacon’s extreme headline cannot simply be discarded, yet its own projections show how strongly the answer depends on assumptions about who votes. I read these differences as evidence of an unsettled electorate, not proof that one survey has revealed the final result.

Uncertainties

The largest gap is the absence of a validated account of likely turnout. Engagement is not a ballot cast. Recalled past voting is not a verified electoral record. Better age-specific turnout evidence, comparable likely-voter measures, and repeated surveys using consistent methods would help distinguish real movement from changes in respondent composition.

The remaining campaign can also change the race. Chiang’s organization can mobilize less-engaged supporters, and undecided voters can move toward him. Shen’s engagement advantage and municipal grievances can instead translate into stronger turnout. Neither continued narrowing nor an incumbent rebound is established by the current evidence.

Some details remain weakly verified. Beacon’s turnout transformation was not independently reproduced. Zmedia’s original release was not directly verified. There are no verified precinct canvass totals or comparable fundraising series here to test the mobilization narrative. Crowd size and online attention do not fill those gaps.

These uncertainties support a close forecast rather than a confident call. They leave the saved assessment unchanged: Yes 46.7%, No 53.3%.