Free China Movement

Evidence · Human Rights · Democratic Future

Take Action

English 简体中文 繁體中文 Français Deutsch Español

🔎

Direct Presidential Elections and Party Alternation in the ROC (Taiwan), 1996–2024

Downloads: PDF · DOCX

1. The Question

Can a polity travel from authoritarian rule to regular, peaceful party alternation at the very top of government — and what institutional design makes that alternation repeatable rather than accidental? The ROC (Taiwan) offers the longest continuous test case in the Chinese-speaking world: eight direct presidential elections between 1996 and 2024, with three completed changes of the governing party in 2000, 2008, and 2016 (Taiwan Paper 1).

2. The Subseries’ Position, Cited

The subseries hub — “Taiwan Democratic Development and Electoral History” — defines its scope as the study of party competition, electoral institutions, voter behavior, and democratic consolidation in the Republic of China (Taiwan), from 1996 to the present (hub).

Its anchor study, Taiwan Paper 1, asks why the Kuomintang (KMT) repeatedly lost presidential contests to the Democratic Progressive Party (DPP) between 2000 and 2024, and answers with a multi-factor, phased framework (paper). Its load-bearing findings for this paper:

  • The first central-level change of ruling party came in 2000, when DPP candidate Chen Shui-bian was elected President (paper).
  • Neither single-factor national-identity determinism nor any thesis of irreversible party decline survives the evidence: the KMT won 58.45% in 2008 after the identity shift was already far advanced, and remained dominant in the 2018 and 2022 local elections (paper).
  • Presidential, legislative, and local elections must be analyzed as three distinct levels. Lai Ching-te’s 2024 victory with only 40.05% of the vote, alongside the DPP’s simultaneous loss of its legislative majority, is direct evidence of split-ticket voting across levels (paper).
  • The study’s conclusions cannot be directly applied to the PRC (Mainland): the two differ fundamentally in social structure, institutional conditions, and civil-society development, and any analogy requires a rigorous comparability argument (paper).

3. Evidence: Eight Elections, Three Alternations, 1996–2024

3.1 The constitutional origin

Direct election rests on the Additional Articles of the ROC Constitution. Article 2 provides that the president and vice president “shall be directly elected by the entire populace of the free area of the Republic of China,” effective from the ninth-term election in 1996; candidates register as joint tickets; “the pair that receives the highest number of votes shall be elected”; and the term is four years, with one consecutive re-election permitted (MOJ Laws & Regulations Database).

3.2 The eight elections

  • 1996 (23 March). Lee Teng-hui and Lien Chan (KMT) won 54.00% (5,813,699 votes) with turnout at 76.04% — the first direct presidential election, held during PRC (Mainland) missile tests in the Strait aimed at intimidating the electorate (1996 election record).
  • 2000. Chen Shui-bian and Annette Lu (DPP) won with 39.30% (4,977,697) in a three-way race: James Soong (independent) took 36.84% and Lien Chan (KMT) 23.10%; the pan-blue camp totaled roughly 59.94%; turnout was 82.69% — the first alternation (Taiwan Paper 1).
  • 2004. Chen was re-elected with 50.11% (6,471,970) against the KMT–People First Party joint ticket’s 49.89%; the margin — 29,518 votes, 0.22 percentage points — is the narrowest in the history of direct presidential elections in the ROC; turnout was 80.28% (Taiwan Paper 1).
  • 2008. Ma Ying-jeou and Vincent Siew (KMT) won 58.45% (7,659,014) against Frank Hsieh (DPP) at 41.55%; turnout was 76.33% — the second alternation (Taiwan Paper 1).
  • 2012. Ma was re-elected with 51.60% (6,891,139); Tsai Ing-wen (DPP) took 45.63% and James Soong (PFP) 2.77%; turnout was 74.38% (Taiwan Paper 1).
  • 2016. Tsai Ing-wen and Chen Chien-jen (DPP) won 56.12% (6,894,744); Eric Chu (KMT) took 31.04%; turnout fell to 66.27% — the third alternation (Taiwan Paper 1).
  • 2020. Tsai was re-elected with 57.13% (8,170,231) — the highest raw vote total for any candidate since direct elections began; Han Kuo-yu (KMT) took 38.61%; turnout was 74.90% (Taiwan Paper 1).
  • 2024. Lai Ching-te and Hsiao Bi-khim (DPP) won with 40.05% (5,586,019); Hou Yu-ih (KMT) took 33.49% and Ko Wen-je of the Taiwan People’s Party (TPP) 26.46%; turnout was 71.86%. The DPP’s third consecutive presidential victory coincided with a hung legislature (KMT 52 seats, DPP 51, TPP 8 of 113) (Taiwan Paper 1).

All 2000–2024 figures above are taken from tables rechecked against the Central Election Commission election database (db.cec.gov.tw), as published in Taiwan Paper 1 (paper).

3.3 What the sequence shows

Three structural facts survive any single-election narrative. First, plurality winners without majorities are a recurring feature of the system: 39.30% in 2000 and 40.05% in 2024 (Taiwan Paper 1). Second, turnout has trended down from its 82.69% peak in 2000 (Taiwan Paper 1). Third, identity transformation is real but not determinative: NCCU Election Study Center tracking shows self-identification as “Taiwanese only” rising from 17.6% in 1992 to 64.3% in December 2020, while “Chinese only” fell to about 2–3% (NCCU Election Study Center via Taiwan Paper 1) — yet the KMT’s 2008 landslide and the TPP’s 2024 youth breakthrough both cut against identity determinism (Taiwan Paper 1).

4. Comparative Options: How Other Democracies Elect a President Directly

  • Plurality in one round — the ROC (Taiwan) model. The highest vote-getter wins; there is no second round. It is decisive and simple, and it accepts sub-majority presidencies, as 2000 and 2024 show (MOJ; Taiwan Paper 1).
  • Absolute majority in two rounds — France. Since the referendum of 28 October 1962 (62.3% yes, 77.0% turnout), the French president has been elected by direct universal suffrage under the two-round system, requiring a majority of the non-blank ballots — a reform adopted precisely to give the office stronger popular legitimacy (1962 referendum record). It guarantees majority support at the cost of a second round and inter-round coalition bargaining.
  • Indirect election through an electoral college — the United States. The states (plus the District of Columbia, for this purpose) elect the President and Vice President through the Electoral College process (U.S. National Archives). It weights federal units but severs the direct link between the national popular vote and the outcome.
  • Single non-renewable term — South Korea. The 16 December 1987 election was the first direct presidential election since 1971; Roh Tae-woo won with 36.64% under first-past-the-post while the two leading opposition candidates combined for over 55% — a split-election cautionary tale (1987 election record). Under the Sixth Republic constitution the presidential term is five years with no re-election permitted, and the governing camp has since changed hands repeatedly, including 1997 (Kim Dae-jung), 2007 (Lee Myung-bak), and 2025 (Lee Jae-myung, elected after Yoon Suk Yeol’s removal following his martial-law declaration and impeachment) (presidential sequence). It removes incumbent-entrenchment incentives at the cost of forfeiting the accountability of a re-election bid.

5. Recommended Design: Numbered Provisions

1. Keep direct popular election of the chief executive as the non-negotiable legitimacy anchor — one citizen, one vote, joint ticket, on the Additional Articles Article 2 model (MOJ). 2. Fix the victory threshold by explicit constitutional choice — plurality (the ROC model, accepting sub-majority winners) or absolute-majority two-round (the French model). Do not leave the threshold to implication. 3. Four-year term, maximum one consecutive re-election — the ROC balance between accountability and anti-entrenchment; the single non-renewable term (Korea) is the stricter variant (MOJ). 4. Independent, permanent election administration with a public results database — the CEC model: official results published to the public database (db.cec.gov.tw); recount and dispute procedures codified before they are needed. 5. A fixed constitutional transfer timetable — the outgoing administration yields power on inauguration day regardless of the margin; disputes go to courts and election commissions, never to force. 6. Design for divided government — when the presidency and the legislature diverge, as in 2024 (DPP presidency, hung legislature), the system must let minority or coalition governance function without extra-constitutional workarounds (Taiwan Paper 1). 7. Opposition-coordination neutrality — the rules neither punish nor reward pre-election coalitions; the split elections of 2000 and 2024 show the costs fall on the splitters, which is where they belong (Taiwan Paper 1). 8. Identity-neutral franchise — the system must work identically whatever the electorate’s identity mix; identity is a background condition, not a design input, per Taiwan Paper 1 (paper). 9. Comparative-use discipline — lessons from the ROC (Taiwan) record must not be mechanically applied to the PRC (Mainland) without a rigorous comparability argument, per Taiwan Paper 1 (paper).

6. Open Questions and Risks

  • Turnout decline. From 82.69% (2000) to 71.86% (2024), with a 66.27% trough in 2016: normalization after democratization’s peak, or civic disengagement? Individual-level survey data is needed before drawing conclusions (Taiwan Paper 1).
  • Sub-majority mandates. Winners at 39.30% and 40.05% governed without majorities; the system provides no runoff. Whether this weakens mandates is contested and unresolved (MOJ; Taiwan Paper 1).
  • Third-party durability. The TPP’s 2024 breakthrough (26.46% of the presidential vote, 8 legislative seats) may mark a structural three-party system or a one-cycle protest vote; a single election cannot decide (Taiwan Paper 1).
  • Cross-strait pressure as confounder. Rising PRC (Mainland) military pressure coincided with DPP victories in 2016–2024, but temporal coincidence is not causation — candidate quality, economic performance, and governing approval moved in the same windows (Taiwan Paper 1).
  • The identity mechanism. “Taiwanese only” identity exceeds 60% while DPP party identification sits at 20–30%; the causal path between the two is underspecified (Taiwan Paper 1).
  • Source limitation. The 1996 figures are drawn from a secondary encyclopedia entry citing the CEC; the drafter verified the CEC database loads but did not re-extract the 1996 figures through its Chinese-language query interface. See the QA notes.

7. Cross-Links