Duverger case study: Taiwan, 2000–2004 — The Cleanest Case on Earth

Duverger's Law predicts that Single Choice Voting in single-winner elections will collapse a polity into two dominant camps, and will mechanically punish whichever camp divides its vote. The Republic of China, Taiwan, gives the cleanest modern demonstration of that effect anywhere on earth — playing out, as the canonical United States elections of 1836 and 1840 once did, across two presidential elections four years apart.

In 2000 the long-governing Nationalists — the Kuomintang, in power since the Republic of China came to Taiwan — divided their vote between two candidates and lost the presidency to the challenger, Chen Shui-bian of the Democratic Progressive Party, who won on 39.30%. It was the first transfer of power in Taiwan's history. Four years later the same camp, reunited behind a single ticket, came within 0.22% — fewer than thirty thousand votes out of nearly thirteen million — of taking the presidency back. The arithmetic is the whole point: a divided majority is beaten by a unified minority, and a camp that consolidates can erase a twenty-point gap almost overnight. The method does not reward the candidate most voters want; it rewards the side that most successfully suppresses its own internal choice.

What followed was eight years of grinding instability, usually laid at Chen Shui-bian's feet — his temperament, his combativeness, the corruption case that later ended his career. Regardless, Chen held a presidency won on a minority of the vote and faced a legislature permanently controlled by the rival camp — a configuration plurality voting made not merely possible but likely. Any president in that position would have governed in crisis. The root cause of the disease has never been properly stated. The cure would have been to let voters rate the whole field instead of forcing them to pick only one. Tragically, that lesson has never been learned, and the solution is still not being put into practice — anywhere in the world.

The race is usually remembered as a three-way contest. That framing erases two structurally important phenomena: the collapse of a serious fourth candidate under tactical-voting arithmetic, and a fifth candidate who openly told his own supporters not to vote for him. One election displayed almost the entire Duverger Syndrome in real time.

The spoiler effect at the top

The two heirs to the ruling Kuomintang together polled nearly 60%. Either one alone would have defeated Chen Shui-bian decisively. James Soong, the popular former provincial governor, lost the Kuomintang's nomination to its official candidate, Vice-President Lien Chan, and ran as an independent; the party expelled him in November 1999. Lien refused to step aside despite trailing Soong consistently in the polls. The governing party's split delivered the presidency to a challenger with 39% of the vote. This is the classical spoiler dynamic in its purest modern form — two candidates from one camp dividing a clear majority and losing to a single rival.

Strategic imprisonment below the threshold

Hsu Hsin-liang was not a fringe figure. He was a former Chairman of the very party that won the election. He ran on a principled cross-strait reconciliation platform, threading between the independence-leaning and status-quo camps. His genuine support was systematically crushed by what the Duverger Syndrome calls strategic imprisonment: voters who found his platform attractive calculated that a vote for him was structurally wasted, and abandoned him for a viable choice. The arithmetic of single-choice voting punished sincere preference. The result was 0.63%. Under a scored ballot, those voters could have rated Hsu honestly alongside their strategic choice, and his real support would have registered. It is one of the cleanest examples of strategic imprisonment in the entire case-study record.

The self-aware non-candidate

Li Ao, standing for the New Party, used the televised presidential debates to tell his own supporters to vote for James Soong instead of him, so that the pro-unification vote would not be split further. He said openly that he was running not to win but to "educate" voters and expose what he considered the dark side of Taiwanese politics. Even the candidate himself understood that the system made sincerity irrational. The result was 0.13%. When candidates campaign against their own election, the voting method has stopped measuring what voters want and started dictating what they may safely express.

The Duverger blind spot

What the campaigns saw mattered as much as what the electorate did. The two Nationalist contenders focused entirely on each other. Lien Chan and Soong fought an intra-party war — the Kuomintang pursued Soong over the Chunghsing Bills affair (興票案) in December 1999; Soong fought back by implicating the sitting president. Both campaigns operated on the shared assumption that Chen could not win and that the presidency would be settled between the two of them. Kuomintang-aligned state media trained its fire on Soong rather than on the eventual winner. Chen's viability was systematically underestimated by both Nationalist campaigns — the spoiler dynamic inverted in plain sight, in which the two largest actors are so fixed on each other that the plurality winner slips through unremarked.

There is a darker reading of the same facts, widely believed among Kuomintang supporters afterwards: that President Lee Teng-hui — the Republic of China's first Taiwan-born president — quietly preferred Chen and threw his weight behind Lien, the weaker of the two contenders, to guarantee the split. Lee was expelled from the party in 2001 and his followers founded the pro-independence Taiwan Solidarity Union, formally joining the opposing camp. Whether the split was engineered or merely structural, the effect was identical: under single-choice voting, a single well-placed actor can determine an outcome by managing who divides whom. The voting method hands that weapon to anyone willing to use it.

The authoritarian neighbour intervenes

Three days before the vote, the Premier of the People's Republic of China warned Taiwanese voters not to "act on impulse" and to shun a pro-independence candidate, with the now-famous declaration that "no matter who comes into power in Taiwan, Taiwan will never be allowed to be independent." The threat was credible: four years earlier, the PRC had fired missiles into the waters around the island during Taiwan's first direct presidential election, and many voters genuinely feared that electing Chen could ignite open conflict. The intervention backfired — it pushed undecided voters toward Chen — and Beijing changed tactics, falling silent four years later. But the structural point is permanent: an external authoritarian power explicitly intervened in an election it could not openly join. Taiwan's Duverger Syndrome operates under conditions of constant existential pressure that compound every one of its symptoms.

The mechanism is unforgiving, and the Nationalist camp responded the only way it could: it consolidated. Lien Chan and James Soong — rivals in 2000 — ran on a single joint ticket, announced on 14 February 2003, more than a year before the election. The same Duverger arithmetic; the opposite strategy. By now the two camps had hardened into the labels Taiwan still uses: the Nationalist-led pan-blue, leaning toward closer ties with Beijing, and the DPP-led pan-green, leaning toward a distinct Taiwanese identity.

The arithmetic tells a story of unity that came too late. Combined, the Nationalists' 2000 vote — 59.94% against Chen's 39.30% — was a commanding twenty-point lead, surrendered only because they ran divided. By the time they finally united, in 2004, the joint ticket could manage just 49.89% and lost by a hair. In four years they had turned a twenty-point lead into a narrow defeat — and the reason cuts deeper than vote transfer. In 2000, many had stayed with the Nationalists out of fear that an openly pro-independence president would provoke the PRC into open conflict. Chen won; no conflict came; Beijing, its threats having backfired, stayed silent. By 2004 that fear had drained away — and with it the camp's central asset. The unified ticket arrived four years too late: the electorate was no longer afraid to re-elect a pro-independence president, and consolidating the vote could not manufacture a fear that events themselves had dispelled. The 337,297 invalid ballots — triple the 2000 figure, swelled by an organised spoiled-ballot campaign — are themselves a Duverger symptom: voters with no third option finding no way to register dissent except by destroying their own vote.

The 228 Hand-in-Hand Rally — a campaign moment that revealed the structure

On 28 February 2004, three weeks before the election, roughly two million citizens joined hands in a 500-kilometre human chain from Keelung to Pingtung, modelled explicitly on the 1989 Baltic Way (itself a chain of some two million people across Estonia, Latvia and Lithuania, formed to protest Soviet rule). The rally commemorated the 228 Incident — the 1947 massacre of Taiwanese civilians — and protested the missiles the People's Republic of China had aimed at the island.

The Kuomintang refused to take part. Facing the question of whether to stand with its political opponents against a foreign military threat, three weeks before a presidential election, the calculation was simple: any action that might help the other camp must be opposed, even at the cost of declining to stand with citizens against missiles pointed at their own homes. This is the earliest publicly visible Taiwanese instance of the symptom the framework calls democracy itself as a wedge issue — the moment defending the polity becomes a partisan position rather than a shared premise — and one of the starkest displays of party over country in the record.

The 319 shooting — a complication, not a refutation

On 19 March 2004, the day before the vote, Chen Shui-bian and Annette Lu were both shot during a campaign motorcade in Tainan. Both injuries were superficial. The pan-blue camp later challenged the result on the theory that the shooting had generated sympathy votes; the lawsuits failed, and the judicial recount in May 2004 confirmed the outcome with the margin slightly reduced. The shooting belongs in the record, but it is a complication around the structural argument, not a refutation of it. Even granting the sympathy effect its maximum plausible weight, the underlying pattern holds: in 2000 a divided camp lost; in 2004 a united one came within a whisker and still fell short, for reasons that ran deeper than any single day's events.

The most immediate casualty of Duverger's Law is the one least often named: the voter's choice itself. In 2000, Taiwanese voters faced five presidential tickets spanning a real spectrum — the two rival heirs of the governing party, Chen's challenge, Hsu Hsin-liang's cross-strait reconciliation, and Li Ao's provocations. Four years later that field had collapsed to two: a single pan-blue ticket against a single pan-green one. Three of the five options had simply vanished from the ballot.

This contraction is not incidental to the system; it is the system working as designed. A democratic election is meant to be the citizens' primary instrument for shaping how they are governed — a sincere chance to express what they actually want. Single Choice Voting hollows that instrument out. It forces the entire electorate into two narrowly defined camps and then asks each voter not "whom do you want?" but "which of these two sides will you pick?" Genuine preference — the third option, the qualified endorsement, the candidate one admires but judges unelectable — is squeezed out before the ballot is even printed. Voters are not so much given a voice as handed a binary, and told that anything outside it is a wasted vote.

That is the deepest sense in which the 2000–2004 pair is a tragedy. It is not merely that a divided camp lost and a united one nearly won. It is that a young democracy, in the space of a single electoral cycle, watched its own range of choice contract by more than half — and mistook the contraction for maturity.

Chen Shui-bian entered office in 2000 with 39.30% of the vote — a plurality, never a majority — and although he won an outright majority in 2004, he never once commanded the legislature. That is the fact that shaped the entire arc of his eight years, and through them Taiwan's institutional development for decades. The friction that became the dominant register of both terms is too readily blamed on Chen himself: divisive, polarising, later prosecuted. The root cause of the friction is structural. The chaos was the output of Single Choice Voting in a two-camp system, not the product of one president's character. Any president who took the executive while the rival camp held the legislature would have governed in crisis. The role produced the deadlock; the personality merely coloured it.

The distinction is not academic. Read as a personal failing, the conclusion is "elect better presidents." Read as a structural output, the conclusion is "redesign the voting method" — which is precisely what would empower the electorate to elect better public servants.

Small Ruling Party, Big Opposition (朝小野大)

The phrase 朝小野大 (cháo xiǎo yě dà — "small in government, large in opposition") described Taiwan's configuration throughout both of Chen's terms. The executive belonged to one camp; the legislature was always controlled by the other camp. The pan-blue alliance held a continuous, hostile majority in the Legislative Yuan from 2000 to 2008 — a majority able to block, frustrate, and weaponise virtually every initiative the executive proposed. The empirical record of that obstruction is consistent:

• Cabinet instability. Six premiers in eight years, the first of them — a respected opposition figure who led a short-lived cross-camp cabinet — gone within five months over the fourth nuclear power plant dispute. Each successor governed under permanent legislative siege.
• Defence procurement blocked. The pan-blue majority blocked a long-pending arms package from the United States — submarines, anti-submarine aircraft, missile defence — rejecting it at the legislature's procedure committee dozens of times across 2004–2006. When the same party returned to the presidency in 2008, it promptly resumed lobbying Washington for the very items it had blocked. Taiwan's defensive posture against an external threat was directly and lastingly weakened by domestic legislative action — the textbook demonstration of party placed over country.
• Appointments frozen. The majority blocked or delayed the executive's nominees to the Control Yuan — the ROC's independent watchdog branch — the Examination Yuan, and the constitutional bench, leaving these oversight bodies under-strength. This was the first run of an appointment-blocking tactic that the constitutional court would later rule unconstitutional — and that the same camp would run again, at greater scale, two decades later.

This is manufactured ungovernability in operation: what the rational strategy of an opposition becomes under single-choice voting when the executive is held by the other camp. Govern badly enough on the other side's behalf, and the next election rewards you. The incentive is structural, and it is exactly backwards.

The aftermath: the binary outlives the term

After the 2008 transition of power, Chen Shui-bian was prosecuted, convicted, and imprisoned on corruption charges, later released on medical parole and partially pardoned. The case remains among the most contested in Taiwan's history: to one side a legitimate exercise of accountability across party lines, to the other a politically driven prosecution of a defeated opponent.

Whichever reading is right, the structural point holds. In a two-camp system hardened by Single Choice Voting, the political class has a standing incentive to turn the prosecution of a former president into a tribal weapon — whether or not the underlying conduct was genuinely criminal. Even a wholly justified prosecution becomes, in such a system, partly an instrument of the ongoing war between the camps. The binary does not end on election night; it follows the loser out of office. A voting method that rewarded broad acceptability over tribal mobilisation would dampen that incentive — without ever shielding genuine corruption from the law.

The 2000–2004 sequence is one of the cleanest Duverger pairs in modern democratic history. Set beside the United States sequence of 1836–1840, the mechanism is identical across two centuries and two continents:

The differences sharpen the point. The Taiwan case carries higher stakes per cycle: an external authoritarian power with explicit annexation intent and visible military pressure. It moved faster — the forced consolidation took four years, not a generation. And it unfolded in an electorate already decades-experienced with single-choice voting, where the American case was a first encounter with the mechanism. In both, the divided side was driven to consolidate. In neither did the polity ask the deeper question: should the voting method have produced this dynamic at all? The Knowledge-Application Gap operates identically across two hundred years — the mechanism is understood well enough to be worked around, and its root is never addressed.

Most case studies illustrate one or two symptoms of the Duverger Syndrome. The 2000–2004 pair displays nine of the twelve at once, several of them in textbook clarity.

Two symptoms are absent because they unfold over longer cycles than a single election pair: the Good-Candidate Drought (3) and the Policy Pendulum (9). Their roots are visible in 2000–2004, but their full expression belongs to Taiwan's later decades.

The Taiwan case carries a feature absent from every other Duverger study: an external authoritarian power that explicitly intervenes in the elections it cannot openly join. In 2000 the People's Republic of China tried open threats, and they backfired. By 2004 it had fallen silent, watching the pan-blue camp reconsolidate around cross-strait engagement. That shift — from external pressure to patient internal exploitation — is the founding moment of the strategy now named in Taiwanese discourse as Nei-tung (內統; nèi tǒng): the project of letting a fractured democracy disable itself from within, working through its political structures rather than against them.

The seeds were planted in this four-year window. A domestic bloc whose cross-strait policy aligned with the threat was forged out of the post-2000 pan-blue consolidation. The template of legislative obstruction under divided government (朝小野大) was rehearsed against the Chen administration. And the redefinition of self-defence as a partisan rather than a national commitment — first visible at the 228 rally — gave an external power something to work with. None of this required an invasion. Single-choice voting supplied the fracture; the authoritarian neighbour learned, patiently, to widen it.

This is also why the case matters beyond Taiwan. Nei-tung is generalisable: any authoritarian state with strategic patience and a target democracy whose binary contains a sympathetic faction can run a version of the same play. The seeds are documented here; the full bloom belongs to the third article in this series, on the 2024 constitutional crisis.

Every visible symptom of the 2000 election dissolves under Informed Score Voting — a scored ballot on which voters rate every candidate across a range rather than choosing exactly one:

• The Soong–Lien spoiler disappears. Voters score both Soong and Lien positively without dividing their support. The stronger candidate emerges from that camp cleanly, with no intra-party war and no expulsion.
• The Hsu Hsin-liang collapse never happens. Voters who genuinely favour his reconciliation platform score him honestly alongside their strategic choice. His real support registers instead of evaporating into 0.63%.
• Li Ao's tactical instruction becomes pointless. He runs on his ideas; voters score him on those ideas; the system reports how many actually agreed, rather than forcing him to campaign against his own candidacy.
• The Duverger blind spot closes. With the whole field scored continuously, there is no "real winner invisible to the dominant actors." There is also no split left to engineer — the single weapon that a well-placed insider could wield under single-choice voting simply has no purchase.
• The divided-government trap (朝小野大) loosens. A scored ballot rewards broad acceptability over tribal mobilisation, producing legislators elected for being widely tolerable rather than narrowly loyal. Cross-camp cooperation becomes structurally available — not guaranteed, but possible — and the manufactured-ungovernability template that defined eight years, and that Nei-tung later inherited, stops being the natural output of the system.

What Informed Score Voting would have changed is not primarily who won — Chen might plausibly have won under any fair method — but the shape of what followed. It would have dissolved the whole split-then-scramble-to-consolidate convulsion of 2000–2004, and produced a legislature in which holding the presidency meant governing rather than waging war by other means. Taiwan's electoral history would not read as a Duverger's Law textbook. It would read as an ordinary democratic record. The Duverger Syndrome is not nature. It is design — and design can be redrawn, before the next external threat finds a Taiwan still arguing about which faction it belongs to.