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

> As of 4 November 2026, Prolepsia puts the chance of yes at 49%.

- Forecaster: Prolepsia (https://prolepsia.com)
- Forecast made: 2026-11-04
- Question type: Yes / no
- Resolves: 2026-11-29
- Status: Open
- Forecasts: 3
- Page: https://prolepsia.com/forecasts/nimbus-calculus-parallax/will-chiang-wan-an-win-the-next-taipei-mayor-election
- Open in the app: https://app.prolepsia.com/library/0b7487d8-89cb-4c9d-911e-660506f27301

## Forecast

- Yes: 49%
- No: 51%

## How Prolepsia reasons

## TL;DR
Chiang Wan-an’s probability of winning is 48.6%; the probability of No is 51.4%. Treat this as a toss-up with a slight lean against reelection, not a safe incumbent race. The decisive issue is which surveys represent the voters who turn out, rather than whether a dramatic late reversal occurs.

## Context
As of November 4, 2026, Chiang faces a competitive reelection contest against Shen Po-yang. Six candidates registered, and the registered field contains no TPP-nominated candidate. That makes the choices of former Huang Shan-shan voters central to understanding the race, but it does not establish where those voters will go. ([Candidate information](https://info.cec.gov.tw/vote2026/question/))

Voting is scheduled for November 28. The question concerns who becomes mayor as a result of that election; an interim appointment does not count. Under the market rules, an election still unresolved by June 30, 2027, 11:59 PM ET resolves to “Other.” Ambiguity is ultimately settled by official Central Election Commission results. ([Election schedule](https://web.cec.gov.tw/api/file/5c5b2a09-2d16-4a13-8445-8a2bdd457221.pdf); [official results authority](https://db.cec.gov.tw/))

## 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 in 1998; Ma Ying-jeou won in 2002; Hau Lung-bin won in 2010; and Ko Wen-je won in 2018. This supports a modest incumbency advantage, not a presumption that current contrary evidence is wrong. ([Taipei historical record](https://www.gov.taipei/cp.aspx?n=C1304E1985BE745E))

Chiang’s own starting coalition also needs care. In 2022 he won 575,590 votes, or 42.29% of valid votes. Chen received 434,558 votes, or 31.93%, and Huang received 342,141 votes, or 25.14%. Turnout was 67.70%. **Chiang inherited a plurality, not a majority.** A different candidate field changes the coalition arithmetic. ([CEC-derived election results](https://vote.nccu.edu.tw/cec/vote3.asp?pass1=L%3A8%3A%3AI%3E%3B88888888iii%28%28))

Current polling is more informative than that historical advantage, but it gives opposing answers. The table includes the earlier observations needed to see the available within-pollster histories, not just the latest favorable or unfavorable headline. All figures are respondent support shares, not win probabilities.

| Organization | Fieldwork or reported period, 2026 | Chiang | Shen |
|---|---|---:|---:|
| TVBS | May | 58% | 30% |
| TVBS | Early October | 49% | 41% |
| TVBS | October 31 | 46% | 42% |
| Hsinmin/Beacon | September 16–18 | 41.7% | 40.5% |
| Hsinmin/Beacon | October 14–16 | 33.3% | 50.3% |
| Z.Media/Shanshui | June 21–22 | 55.2% | 27.5% |
| Z.Media/Shanshui | October 18–19 | 39.0% | 42.8% |
| UDN | October 20–22 and 25 | 41% | 45% |
| ETtoday | October 21–25 | 48.6% | 41.1% |
| ETtoday | October 31–November 1 | 52.0% | 40.7% |
| ETtoday | November 1–3 | 50.7% | 41.9% |
| Strategy Institute/Wowo | October 31–November 1 | 35.3% | 52.6% |

The histories show substantial narrowing in TVBS, Hsinmin and Z.Media, while ETtoday continues to show a Chiang lead. Wording continuity between Z.Media’s June and October surveys was not fully verified, so that comparison is less clean than it looks. ([TVBS history](https://news.tvbs.com.tw/politics/4044368); [Hsinmin report](https://dc579968-5ed9-4902-8454-9871fa44324e.filesusr.com/ugd/3ea6ac_257951a3d235492f8595f6dcd7b5cdd3.pdf); [Z.Media report](https://www.zmedia.com.tw/Document/PoolDetail/44789); [UDN](https://udn.com/news/story/124652/9840910); [ETtoday reporting](https://n.yam.com/Article/20261104740155); [poll comparison](https://udn.com/vote2026/pollboard); [Wowo report](https://www.wwmt.com.tw/polls/202610311101/))

The central conflict is between substantial surveys, not merely between a good poll and an obvious outlier. TVBS completed 1,747 live-landline interviews on October 31; its headline voting intention uses 1,666 respondents willing to vote. UDN completed 1,115 landline/mobile interviews during October 20–22 and 25. They show opposite four-point leads. Their different coverage and voting-intention denominators matter. ([TVBS original report](https://www-asset.tvbs.com.tw/uploads/opinion-polls/1554/1793759962_1510311___________________________________________________.pdf); [UDN report](https://udn.com/news/story/124652/9840910))

ETtoday adds recent evidence favorable to Chiang. Its November 1–3 survey collected 1,853 responses through an invited online questionnaire drawn from its membership database. That is evidence of support within a different sampling frame. It is not a decisive rebuttal to the telephone surveys, because membership recruitment and willingness to respond can create errors outside the published sampling margin. ([ETtoday methods and results](https://n.yam.com/Article/20261104740155))

Recency cannot explain away the disagreement. Over the same October 31–November 1 window, ETtoday showed Chiang ahead 52.0% to 40.7%, while Wowo showed Shen ahead 52.6% to 35.3%. The candidate-margin gap between those surveys was 28.6 percentage points. Ordinary sampling noise is not an adequate account of that difference. ([Contemporaneous comparison](https://newtalk.tw/news/view/2026-11-02/1068760))

The adverse evidence also needs qualification. Hsinmin/Beacon’s raw 33.3%–50.3% result is not the same thing as its election projection. Reported turnout-and-past-vote scenarios put Chiang at 46.9%–49.9% and Shen at 49.2%–52.9%, a much closer contest. These are conditional projections, not confidence intervals. The transformation was not reproducible from a complete published specification, so neither the raw headline nor the adjusted scenarios deserve automatic priority. ([Reported electorate scenarios](https://www.businesstoday.com.tw/article/category/183027/post/202610200011/))

Wowo’s AI telephone survey interviewed people, but its documentation leaves unresolved questions about dual-frame overlap and number generation. Its reported demographic weighting changes the result only slightly, so ordinary demographic adjustment alone does not explain its large Shen lead. Those issues limit its force without establishing fabrication. A separate TPOC release uses 17,550 simulated AI citizens rather than interviewed voters; it provides no independent measurement of voter preferences. ([Wowo methodology](https://www.wwmt.com.tw/polls/202610311101/); [TPOC release](https://www.tpoc.com.tw/ArticlePage/879ad2b3-68ff-472f-b9d4-dae51b61e066))

The strongest warning beyond the headlines is coalition leakage. UDN finds recalled Huang voters supporting Shen 55% to 38%; TVBS reports a similar 53% to 39% split. Yet current TPP identifiers strongly favor Chiang: 75% to 16% in UDN and 80% to 14% in TVBS. These groups are not interchangeable. The absence of a TPP candidate does not deliver Huang’s entire former constituency to Chiang. ([UDN transfer findings](https://udn.com/news/story/124652/9840910); [TVBS tables](https://www-asset.tvbs.com.tw/uploads/opinion-polls/1554/1793759962_1510311___________________________________________________.pdf))

Participation signals also favor caution about Chiang. UDN reports election interest among 91% of Shen supporters and 77% of Chiang supporters. Z.Media finds Shen ahead 51.2% to 38.6% among definite voters, compared with 42.8% to 39.0% overall. These are self-reports, not validated turnout. They identify a risk; they do not establish the election-day electorate. ([UDN](https://udn.com/news/story/124652/9840910); [Z.Media](https://www.zmedia.com.tw/Document/PoolDetail/44789))

## What's non-obvious
The tempting story is that Chiang has a lead and Shen needs a late surge, or that Shen’s momentum has already made Chiang the clear underdog. Neither reading fits the evidence. The disagreement is about the electorate now. A Chiang victory can follow from the favorable surveys being more representative, just as a Shen victory can follow from coalition leakage and participation differences already visible in other surveys.

Past polling successes do not settle which method is right today. The available historical checks cover selected elections, with changes in collection methods, sponsors and timing. They do not establish a stable correction for any organization. Debate audiences are similarly easy to misread: TVBS finds viewers favoring Shen and nonviewers favoring Chiang, but those self-selected groups do not measure persuasion caused by watching. ([Historical polling discussion](https://news.tvbs.com.tw/politics/1962665); [debate findings](https://news.tvbs.com.tw/politics/4044014))

## Uncertainties
The largest gap is political representativeness. There is no harmonized microdata comparison or comparable nonresponse study that explains the opposing results. Demographic balance does not prove political balance. Recalled past votes also cannot be treated as verified individual voting records.

Actual turnout and undecided allocation remain unresolved. Supporter enthusiasm can matter, but the evidence does not justify converting every stated intention into a ballot or assuming that higher turnout automatically favors one party. Campaign movement remains possible before November 28, but extending earlier polling declines in a straight line would add an unsupported assumption. ([Official election date](https://web.cec.gov.tw/api/file/5c5b2a09-2d16-4a13-8445-8a2bdd457221.pdf))

The most useful additional evidence would be fresh, independently sponsored surveys with full questionnaires, clear sampling and turnout definitions, and transparent treatment of nonresponse and overlapping phone frames. Comparable evidence across methods would help distinguish real movement from different survey populations. Until that gap closes, the forecast remains 48.6% Yes and 51.4% No.

## What changed since the last forecast

Between the October 31, 2026 and November 4, 2026 runs, the forecast barely moved (p(yes) 0.47 → 0.49 (+1.9pp; odds shift 1.1×)). No meaningful change.

## Forecast history

| Date | Forecast |
|---|---|
| 2026-11-04 | 49% chance of yes |
| 2026-10-31 | 47% chance of yes |
| 2026-10-27 | 48% chance of yes |

## Sources Prolepsia read

- [https://info.cec.gov.tw/vote2026/question/](https://info.cec.gov.tw/vote2026/question)
- [Election schedule](https://web.cec.gov.tw/api/file/5c5b2a09-2d16-4a13-8445-8a2bdd457221.pdf)
- [https://www.gov.taipei/cp.aspx?n=C1304E1985BE745E](https://www.gov.taipei/cp.aspx?n=C1304E1985BE745E)
- [選舉概況](https://vote.nccu.edu.tw/cec/vote3.asp?pass1=L%3A8%3A%3AI%3E%3B88888888iii%28%28)
- [https://dc579968-5ed9-4902-8454-9871fa44324e.filesusr.com/ugd/3ea6ac_257951a3d235492f8595f6dcd7b5cdd3.pdf](https://dc579968-5ed9-4902-8454-9871fa44324e.filesusr.com/ugd/3ea6ac_257951a3d235492f8595f6dcd7b5cdd3.pdf)
- [https://udn.com/news/story/124652/9840910](https://udn.com/news/story/124652/9840910)
- [〖ET民調〗蔣萬安支持度50.7％領先沈伯洋41.9％　市政解決能力不分上下 \| 蕃新聞](https://n.yam.com/Article/20261104740155)
- [https://www.wwmt.com.tw/polls/202610311101/](https://www.wwmt.com.tw/polls/202610311101)
- [https://www-asset.tvbs.com.tw/uploads/opinion-polls/1554/1793759962_1510311___________________________________________________.pdf](https://www-asset.tvbs.com.tw/uploads/opinion-polls/1554/1793759962_1510311___________________________________________________.pdf)
- [https://www.businesstoday.com.tw/article/category/183027/post/202610200011/](https://www.businesstoday.com.tw/article/category/183027/post/202610200011)

## The question

Imported from Polymarket.
This yes/no question comes from a single Polymarket market.

### How it resolves

Taiwanese local elections are scheduled to be held on November 28, 2026.

This market will resolve according to the candidate who becomes the next Mayor of Taipei as a result of this election.

Temporary, interim, or placeholder Mayors appointed before the election will not be considered.

If the result of this election isn't known, or no candidate has been elected, by June 30, 2027, 11:59 PM ET, the market will resolve to "Other".

This market will resolve based on the results of the elections as indicated by a consensus of credible reporting. If there is ambiguity, this market will resolve based solely on the official results as reported by the Taiwanese government, specifically the Central Election Commission (https://db.cec.gov.tw/).

---

Prolepsia’s forecasting engine made this forecast: it researched the question, weighed the evidence and wrote the report on this page. A probability is not a promise; this one changes as the evidence does.