Duverger case study: Taiwan, 2005 — The Reform That Broke the Parliament

Over twenty years on, almost no one in Taiwan talks about the 2005 reform of the Constitution of the Republic of China. Where it is remembered at all, it is remembered as a dull procedural "modernisation." It was nothing of the kind. It was the single most consequential structural decision in Taiwan's democratic history, and its repercussions run straight to the constitutional crisis of the 2020s.

To a public worn down by a legislature it saw as bloated, brawling and corrupt, the reform was sold as a clean-up: halve the chamber, the promise went, and the politics would improve. The promise was broken. Two decades on the legislature is indeed smaller — and the worst-brawling, most gridlocked in the country's history. Yet it would be wrong to blame the voters for that result. Electoral systems are an arcane subject, and the change that actually mattered was buried in the technical detail. The public was not applauding a reform it understood; it was sold one.

The political parties understood it perfectly — which is the tell. The two largest, the Kuomintang and the Democratic Progressive Party, backed the reform together, while every small party, across the ideological divide, opposed it. For one moment the two giants set their enmity aside and acted as an ad-hoc political cartel, colluding to entrench their own dominance and wall the smaller parties out. The biggest loser in 2005 was not any single party. It was the Taiwanese public — the ROC electorate, whose range of real choice was quietly narrowed for a generation.

What the parties grasped and the public could not see was the mechanism. The reform re-wired the legislature from SNTV — the single non-transferable vote, a system that, for all its faults, let small parties live — to single-member-district mixed-member majoritarianism, the textbook generator of Duverger's Law. That is the structural heart of this article, and it carries a lesson that reaches well beyond Taiwan: cross-party consensus is not always a marker of democratic health. Sometimes — as in 2005 — it is the signature of a cartel.

The Promise: Halve the Bloated, Brawling Legislature

The reform rode a wave of public fury. The seat-halving movement (立委席次減半運動) had turned popular disgust at a Legislative Yuan seen as bloated, brawling and corrupt into a single demand — cut the legislature in half — and in the 2004 campaign both big parties raced to claim the banner. The reform was sold on three promises: end the brawls, cut the corruption, streamline governance.

The instrument was a constitutional amendment. Much like the United States, the Republic of China leaves the text of its original (1947) constitution untouched and instead appends amendments — the Additional Articles of the Constitution of the Republic of China (中華民國憲法增修條文) — which override the original wherever they conflict, just as the US Twenty-First Amendment repealed the Eighteenth. The 2005 package was the seventh revision of those Additional Articles since 1991 (第七次增修). The Legislative Yuan proposed it on 24 August 2004; it was ratified by the only ad-hoc National Assembly in the Republic of China's history — elected at a record-low 23.36% turnout — by 249 votes to 48 on 7 June 2005, and took effect with the legislature elected in 2008. It made five changes:

• The Legislative Yuan was cut from 225 seats to 113 — 73 single-member districts (every county guaranteed at least one), 6 aboriginal seats, and 34 party-list seats allocated to parties clearing a 5% threshold, with women guaranteed at least half of each list.
• Terms were lengthened from three years to four, synchronised with the presidential cycle.
• SNTV was replaced by single-member-district mixed-member majoritarianism (並立制, MMM) — the decisive change, and the subject of this article.
• The National Assembly was abolished, and future constitutional amendments were routed to a national referendum.
• Presidential impeachment was moved to the Constitutional Court (half the legislature to propose, two-thirds to resolve, the Grand Justices to adjudicate).

The Broken Promise: In 2024, the Worst Brawl Ever

Judge the reform by its own promise. The brawls did not end: on 17 May 2024 the Legislative Yuan staged the most violent physical fight in its democratic history, with several legislators carried out and hospitalised. Corruption did not vanish; it migrated. And governance was not streamlined; the chamber descended into the worst gridlock on record. Point for point, the populist clean-up delivered the opposite of what its supporters had wanted.

It is one of the cleanest cases anywhere of a legible-but-wrong reform beating a correct-but-complex one. "There are too many of them — cut them" is satisfying, easy to grasp, and easy to vote for. It was also exactly the wrong fix, and the chamber it built proved it twenty years later.

The Cartel Moment: When Blue and Green Agreed

Ratifying the reform produced the single most revealing election in modern Taiwanese politics: the one occasion when the Blue–Green binary visibly dissolved. It did not dissolve into unity. It dissolved into duopoly versus pluralism. The two big parties — the Democratic Progressive Party and the Kuomintang — lined up together for the reform. The small parties of both camps — the pan-green Taiwan Solidarity Union and the pan-blue People First Party and New Party — lined up together against it. Sworn cross-strait enemies found themselves on the same side, because the interest at stake was not sovereignty but survival.

This is the clearest empirical glimpse Taiwan offers of the two dominant parties behaving as a cartel. The Blue–Green binary is usually presented — by the parties, and by the public conversation — as an unbridgeable cleavage of identity. In 2005 its two poles coordinated perfectly the moment their shared interest in walling out everyone smaller was on the table. The binary was suspended precisely in order to be made permanent. (The detail that seals the reading: the three-quarters ratification threshold was itself only fixed by the legislature the week after the assembly was elected — once it was already clear the pro-reform parties held more than 75% of the seats.)

The word is precise, not rhetorical. Political scientists call this a cartel party dynamic — the term Richard Katz and Peter Mair gave it in 1995 — in which established parties stop genuinely competing and instead collude, using the resources and the rules of the state to entrench themselves and to keep smaller outsiders from ever becoming viable. A political cartel is to a democracy what a price-fixing cartel is to a market: the dominant players quietly agree not to compete where competition would threaten them all. On every ordinary day the KMT and the DPP are mortal enemies; on the one question that threatened them both — whether smaller parties could share power — they fixed the price together.

This is also why no large party will ever propose a reform like Informed Score Voting — and why, if such a reform is ever to come, it will not come from inside the legislature. The binary is not merely an organic feature of Taiwanese society; it is, in part, a shared asset of the two parties that benefit from it — the one asset a better voting method would expropriate. A cartel does not vote to dissolve itself. A better election method will have to be imposed on the cartel from outside it — by the electorate the cartel was built to manage. That is the task this series ultimately points toward: the public, the reform's biggest loser, is also the only force that can undo it.

The Wrong Cure for a Real Disease

The disease was real. SNTV bred factional machine politics and intra-party vote-buying; the reformers were not imagining the problem. But they misdiagnosed it. The culprit was the single-vote constraint — the same flaw that runs through all Single Choice Voting — not the multi-member district, which was the one feature worth keeping. By "curing" SNTV with single-member districts, the reform preserved the single-vote flaw and added Duverger's mechanical two-party squeeze on top, sharpening every symptom of the Duverger Syndrome rather than relieving it. The right direction from SNTV was toward better multi-member elections; the reform went toward simpler single-member ones.

It compounded the error in the detail. Taiwan adopted the non-compensatory variant of the mixed system — mixed-member majoritarian (並立制, MMM; the parallel system used in Japan) — rather than the compensatory mixed-member proportional (聯立制, MMP; used in Germany): under MMM the district and party-list seats are counted separately, so the list cannot correct the distortion the districts produce. With only 34 of 113 seats on the list, and with no dual candidacy to soften the single-member contests, Taiwan runs an unusually unforgiving version of an already disproportional system.

What the Machine Did: The Small Parties Vanish

The result the design predicted arrived within a single cycle. In the first election under the new system, in 2008, the small parties were annihilated: the Taiwan Solidarity Union, the People First Party and the New Party — all of them real blocs under SNTV — were swept from the constituency map. The single-member tier, 73 of the 113 seats, is pure Duverger's Law: a party with around 40% of the district vote, efficiently spread, takes a commanding majority of district seats, and the thin party-list tier cannot make up the difference.

This is the most immediate casualty of the law, turned into machinery. The first of these three Taiwan case studies — Duverger case study: Taiwan, 2000–2004 — watched the presidential ballot contract from five tickets to two; the 2005 reform did the same thing to the legislature — permanently, and by design. It deleted the smaller parties from the district ballot before a single vote was cast. A voter cannot choose what is not on the ballot, and after 2008 the district ballot was, in all but name, a choice of two.

The proof is written into what became of the anti-establishment energy that built up, over the following decade, against both big parties. When that energy finally found a vehicle — the Taiwan People's Party — the new party could win seats only on the proportional party-list vote, the same fate the table shows the Taiwan Solidarity Union meeting in 2012. In the district contests, where 73 of the 113 seats are decided, the new party won none. The single-member tier strangles a new entrant at birth; the narrow list tier barely lets it breathe. What the Taiwan People's Party then did with the seats it held, and whose interests it came to serve, is the subject of the third Taiwan case study, on the constitutional crisis of 2024 onward.

The same 2005 package left two devices in the constitution that lay dormant for two decades and then detonated. The first is the presidential-impeachment procedure — half the legislature to propose, two-thirds to resolve, the Constitutional Court to adjudicate — the exact mechanism aimed at President Lai Ching-te in 2025–2026. The second is a near-impossible amendment bar: a quarter of legislators to propose, three-quarters to resolve, then a referendum requiring yes-votes from more than half of the entire electorate — a threshold no amendment has cleared since (the 2022 proposal to lower the voting age failed at it).

By sealing the constitution, the reform did not end constitutional conflict; it displaced it downward into ordinary-legislation warfare, where the guard-rails are weaker. There is an accidental mercy in the lock — a tribal majority that cannot easily amend the constitution cannot easily amend it for the worse, either — but that is luck, not design, and the durable safeguard against bad amendments was never a frozen constitution; it is an electorate and a legislature that would not pass them. How these dormant devices were fired, and what the locked constitution channelled the conflict into, is the story that third case study tells.

Had the 2005 reform replaced SNTV with Informed Score Voting in multi-member districts instead of single-member majoritarianism, the second half of this story would not exist. The constituency tier would have stayed plural: the Taiwan Solidarity Union, the People First Party, and later the Taiwan People's Party would have won district seats on their genuine support rather than depending on a party-list bailout. The legislature would have mirrored a multidimensional electorate instead of a two-camp map, and the hyper-binary chamber whose deadlock drives the present crisis would never have formed. The impeachment and amendment clauses could have been written word for word and would have mattered far less, because the chamber wielding them would not have been a tribal duopoly.

2005 is the point at which Taiwan could have turned toward consensus democracy and turned, instead, toward a sharper binary — with the public's consent, secured by a promise that was never kept. That is the hinge, and the tragedy, of this case study. The disease was correctly felt. The wrong cure was chosen — by the near-unanimous agreement of the parties it would protect — and twenty years on the patient is sicker for it.