Taiwan: The Court Went First, and the Law Carries Its Name

Taiwan is described, almost always in the same four words, as the first in Asia. The phrase is accurate and the way it is used is not. It arrives in most accounts attached to an implication — that a society reached a settled view, and then its law caught up — and that implication is the opposite of what happened.

What happened is that three institutions disagreed in public over two years, in a sequence that can be dated. A constitutional court held in May 2017 that the Civil Code was unconstitutional in this respect and gave the legislature two years to fix it. In November 2018 the electorate was asked five questions touching on the same subject and answered four of them against the direction the court had set, by margins in the millions. In May 2019 the legislature passed a law that gave same-sex couples marriage registration and almost every incident of marriage, and it did so through a separate statute named after the court ruling rather than by amending the Civil Code the referendum had just voted to protect.

That last move is the essay. A country does not name a marriage law after a docket number by accident. The title of the Act for Implementation of J.Y. Interpretation No. 748 contains neither the word marriage nor the word same-sex. It is the referendum’s fingerprint, left on the statute book, and it is still there.

The chamber asks what becomes visible about lives in Japan when another society is examined closely. Taiwan is worth examining closely for two reasons at once. It is the direct test of the claim that this kind of change requires a Western cultural inheritance, and it fails that claim cleanly. And it is a case in which the gap between a legal outcome and a social consensus was not a matter of inference. It was put to a vote and counted.

The case reached the Constitutional Court by the ordinary route of one person refusing to stop. An activist who had campaigned since the 1980s was refused marriage registration with his partner in 2013, pursued the refusal through the administrative courts, and petitioned for constitutional interpretation in 2015. The Taipei City Government, which had been the body doing the refusing, petitioned as well: it wanted to be told what the law was.

On 24 May 2017 the court announced Interpretation No. 748. It held that the Marriage Chapter of the Civil Code, to the extent that it does not allow two persons of the same sex to create a permanent union of an intimate and exclusive nature for the purpose of living a common life, violates both the freedom of marriage protected by Article 22 of the Constitution and the right to equality protected by Article 7.

Two features of the holding decided everything that followed. The first was the remedy. The court gave the relevant authorities two years to amend or enact law, and said that if they did not, same-sex couples would be entitled to have their marriage registration effectuated at the household registration authorities anyway. That is a deadline with teeth: inaction would produce the outcome, not prevent it.

The second was a sentence of judicial modesty that would later carry a great deal of weight. The court said that it was within the discretion of the authorities concerned to determine the formality by which equal protection of the freedom of marriage was achieved, and it named the options: amending the Marriage Chapter, adding a special chapter to the Family part of the Civil Code, enacting a special law, or some other formality. The court had specified the destination and left the road open. Eighteen months later, a referendum would walk down one of those roads and shut the others.

In December 2017, seven months after the interpretation, the legislature amended the Referendum Act in a direction that had been sought for years on democratic-participation grounds. The second-stage signature requirement fell from five per cent of the electorate to one and a half per cent, roughly from nine hundred thousand signatures to under three hundred thousand. The passage rule became a double threshold: the yes votes must exceed the no votes and must also reach a quarter of all eligible electors. The voting age for referendums came down to eighteen.

The reform was not made with this question in mind, and it is the reason this question could be asked. A coalition organised largely through Christian churches — a striking fact in a society where Christians are under five per cent of the population, and the reason this essay runs partly through the religion lens — gathered the signatures and put its questions on the ballot for 24 November 2018, the day of the local elections.

Five questions are relevant. Case 10 asked whether marriage as defined in the Civil Code should be restricted to a union between one man and one woman: 7,658,008 voted yes and 2,907,429 no. Case 11 asked whether the education ministry and schools should not implement the homosexuality-related instruction prescribed in the enforcement rules of the Gender Equity Education Act at elementary and junior high level: 7,083,379 yes, 3,419,624 no. Case 12 asked whether the rights of same-sex couples living together permanently should be protected by means other than changing the Civil Code: 6,401,748 yes, 4,072,471 no. Case 14, put by the other side, asked whether same-sex marital rights should be protected within marriage as defined in the Civil Code: 3,382,286 yes against 6,949,697 no. Case 15, also put by the other side, asked whether gender equity education including emotional education, sex education and education about homosexuality should be taught at all levels of national education: 3,507,665 yes against 6,805,171 no.

Read the five together and the electorate had said something more specific than no. It had said: not in the Civil Code, and not in the schools. Roughly two and a half times as many people voted to keep the Civil Code definition as voted against it, on a turnout near fifty-six per cent of an electorate of nearly twenty million. No serious reading of those numbers supports the sentence that Taiwan had already agreed.

A referendum in Taiwan cannot overturn a constitutional interpretation, and nobody with standing claimed otherwise. The Referendum Act requires that a passed proposal on legislative principles be answered by a government bill within three months, and it operates inside the Constitution rather than above it. The two-year clock kept running.

So the government drafted to satisfy both instruments at once, which was possible only because of that sentence of judicial modesty in 2017. The court had said a special law would do. The referendum had said not the Civil Code. A special law it was.

On 17 May 2019 the Legislative Yuan passed the Act for Implementation of J.Y. Interpretation No. 748 at third reading; the vote on the central article was 66 in favour and 27 against, with seven legislators from the main opposition party crossing to support it. The president promulgated it on 22 May and it took effect on 24 May 2019, two years to the day after the ruling. Five hundred couples registered before five o’clock that afternoon.

What the Act does is worth stating precisely, because the precision is the compromise. Article 1 says the Act is enacted to enforce Interpretation No. 748. Article 2 does not say marriage: it says that two persons of the same sex may form a permanent union of an intimate and exclusive nature for the purpose of living a common life, which is the court’s phrase, copied out. Article 4 then says that such a union is effected in writing, with two witnesses, by marriage registration at the household registration authority — so the word appears after all, one article later, attached to the act of registering rather than to the thing itself. A later article applies the Civil Code’s provisions on spouses to these unions with the necessary changes. The substance is marriage. The architecture is a separate building with the court’s case number over the door.

This is where Taiwan earns its place in the chamber, because it settles an argument that is made here often and made loosely. The argument holds that this kind of legal change belongs to a particular civilisational package — individualism, a Protestant or post-Christian settlement, a rights vocabulary imported from the Atlantic — and that a society without that inheritance either cannot produce it or produces only a hollow copy. Taiwan is the counter-example that does not need to be argued for. It is a Chinese-speaking society with a Confucian family inheritance, ancestral rites, a living practice of filial obligation, and a religious landscape in which Christianity is a small minority. It got there first in the region, and it got there while its electorate was voting the other way.

So the preconditions were not cultural. What were they? Four things, and they are institutional rather than spiritual. A constitutional court with the standing to strike down the Civil Code and the willingness to use it. A petitioner who had been refused in 2013 and was still going in 2015, plus a city government that wanted the question settled rather than avoided. A remedy with a clock and a default, so that legislative delay produced the outcome instead of postponing it. And a legislature with the arithmetic and the nerve to pass something within the deadline, in the year before a national election, having just been told by seven and a half million people not to.

None of those four is about whether a society is ready. Three of them are about the design of a constitutional system and one is about a particular set of people spending their own credibility. That is the finding, and it is not a comfortable one for either side of the usual argument. It denies the conservative claim that the change is foreign. It also denies the liberal claim that the change follows automatically once opinion has moved, because in Taiwan opinion had not moved when the change was made.

The referendum is usually described as having been overridden. It was not, and the clearest evidence is what happened to the education questions, where there was no constitutional interpretation to protect anyone.

Marriage was delivered because a court had ordered it. Gender equity education had no such order, and there the referendum result held. The enforcement rules under the Gender Equity Education Act were revised, and the explicit reference to instruction about homosexuality gave way to broader wording about respecting different genders, gender characteristics, gender identities and sexual orientations. Teachers and administrators, by the accounts of people who work in Taiwanese schools, became markedly more cautious about the subject regardless of what any rule said. That is the shape of the thing this essay is arguing: the court could deliver the register office and could not deliver the classroom.

The campaign also has a measured cost, and the measurement is the reason this section exists rather than being a matter of sympathy. A study published in 2019 in the International Journal of Environmental Research and Public Health compared two online survey waves of adults in Taiwan, the first in January 2017 and the second one week after the referendum. Among non-heterosexual respondents, significant suicidal ideation was 15.4 per cent in the first wave and 24.6 per cent in the second, while among heterosexual respondents the figure moved from 5.2 per cent to 6.3 per cent. The authors are explicit about the limits: two separate cross-sectional samples recruited through social media rather than a followed panel, a second wave taken at the acute moment, and no mechanism identified.

Set against it is the other half of the record, which is as important and less quoted. A prospective study of 863 gay and bisexual men, surveyed shortly before the law took effect in May 2019 and again in late 2020, found modest but significant reductions in depressive symptoms and in externally sourced minority stress, and more disclosure to friends, family and parents. Its authors are equally explicit about a non-probability sample, thirty-one per cent attrition, no comparison group, and a follow-up wave that fell inside the pandemic. Neither study proves a general claim. Together they describe a country that paid a bill during the campaign and received something after the law.

Two things have happened to Taiwanese opinion since 2019 and they should be reported separately, because the instruments disagree and reporting only the friendlier one is the exact failure this essay is about.

On the government’s own annual gender-equality survey, agreement that same-sex couples should have the right to marry rose from 37.4 per cent in 2018 to 69.1 per cent in 2024. On a survey commissioned by an advocacy organisation and conducted by a polling firm in April 2026, with a sample of 1,077 adults, support for same-sex marriage stood at 54.3 per cent, up about twelve points over seven years, with higher figures for joint adoption and for transnational recognition and the largest single rise on gender equality education. Both series show movement in the same direction. They do not agree on where the country now is, and the gap between a government instrument and an advocacy-commissioned one is not a detail a careful reader should be asked to ignore.

The law has continued to move too, and mostly by closing gaps the 2019 compromise left. In January 2023 the interior ministry instructed local authorities to register marriages between a Taiwanese national and a foreign partner even where the partner’s own country does not recognise such a marriage; partners from mainland China were excluded, because cross-strait family matters run under a separate statute, and that exclusion was itself eased in September 2024 for couples who had registered in a third jurisdiction. In May 2023 the legislature amended the adoption article so that a couple could adopt jointly and could adopt a child who was not the biological child of either, closing an omission that had left some families half-recognised for four years.

And some of it is unresolved, which the essay records as unresolved rather than tidying. Access to assisted reproduction has not been equalised. A Cabinet draft approved in December 2025 would extend eligibility to unmarried women and to female spouses in same-sex marriages, with surrogacy separated out for a different argument; at the time of writing it is before the legislature and is not law.

Japan has never held a vote of this kind, and there is no mechanism by which it could hold one. The only national referendum in Japanese law is the one under Article 96 of the Constitution for ratifying a constitutional amendment, governed by the 2007 Act on procedures for amendment, and it has never been used. There is no equivalent of the Taiwanese instrument by which citizens can put a legislative principle on a ballot. Whatever else that means, it means the Taiwanese experience of standing in a supermarket queue in the autumn of 2018 next to posters arguing about whether your family should exist has no Japanese counterpart.

Recognition has advanced instead along two tracks. The first is municipal. Shibuya and Setagaya began issuing partnership certificates in November 2015, and by the joint count of Shibuya City and a specialist non-profit, as of 31 May 2025 the systems covered 530 municipalities and 92.5 per cent of the population, with 9,836 couples registered. The second track is litigation: six actions brought in five places, framed as claims for state compensation, in which the constitutional question is decided on the way to deciding the claim.

Both tracks have a property worth naming. The certificates have no effect under national law. They do not alter the family register, they do not create inheritance rights, they do not confer the spousal deduction, they do not resolve a partner’s residence status. They are real in a hospital corridor and absent in a probate court, and the gap between those two facts is a thing couples discover at the worst possible time.

The litigation has gone further and has now split. The Sapporo High Court held on 14 March 2024 that the relevant Civil Code and family register provisions are unconstitutional; high courts in Tokyo in October 2024, Fukuoka in December 2024 and Nagoya and Osaka in March 2025 reached findings of unconstitutionality on overlapping constitutional grounds. Then on 28 November 2025 a second Tokyo High Court bench held the opposite, and observed that the matter should be deliberated in the Diet. On 25 March 2026 the Supreme Court referred the cases to its Grand Bench, which is what it does when it intends to settle a constitutional question for the country. At the time of writing that bench has not ruled, and commentary expects a judgment no earlier than 2027. This essay states that as an open question, because it is one.

Set Taiwan beside the case this chamber has already examined in Australia, where a national postal survey in 2017 returned a clear majority and Parliament amended the Marriage Act within a month, and where the campaign’s cost to the people being voted on is also in the journals. Put the two next to Japan and a comparison offers itself that is more useful than the usual one.

The usual comparison ranks the outcomes: who has it, who does not, who is ahead. That comparison is available and it is nearly useless, because it tells a reader nothing she can do anything with and invites the conclusion that her own country is behind, which is a mood rather than a finding.

The comparison this essay proposes is about procedure. In each country the question was eventually decided by whichever institution was willing to move before the others were ready, and each institution, in going first, sent the bill to different people. In Australia the electorate went first, and the bill went to the minority whose family life was the subject of a three-month national campaign. In Taiwan the court went first, and the bill went partly to the same people — the referendum campaign was a direct response to the ruling — and partly to a legislature that had to vote against a fresh popular instruction. In Japan nobody has gone first, and the bill has gone to a small number of named plaintiffs carrying six lawsuits for years, and to everyone else in a slower and quieter form: a certificate that works in some buildings and not others, and a wait whose end is not scheduled.

The essay declines to rank these. A court that moves ahead of opinion is not obviously braver than an electorate that moves; a vote that is won is not obviously cleaner than a ruling that is obeyed. What the three cases share is that there is no version in which nobody pays. The only real variable is who, and how visibly, and whether they chose it.

This house sells private evenings, and it should say against its own interest what that has to do with an essay about constitutional procedure. What it sells is a room where the terms are agreed in advance and are not subject to anybody else’s vote. That is a commercial proposition, and its appeal grows in proportion to how much of a woman’s life is arranged by defaults she did not choose and cannot easily amend. The house benefits from that condition. It did not create it and it does not fix it.

The honest limit is that a private arrangement is the one form of recognition that requires no institution at all, which is both its whole value and the precise reason it cannot substitute for one. Nothing bought for an evening survives a probate court, a hospital admission after hours, or a residence application. The Taiwanese story is about the machinery that does survive those, and about what it takes to move it. A room is not that machinery and should never be described as though it were.

What one evening can honestly offer is smaller and still worth naming: an occasion in which the terms were set by the two people in the room, written down beforehand, and not renegotiated by anyone else. Whether that experience is useful outside the room is not in the house’s gift.

It does not claim that Taiwan is a model. The route it took depended on a constitutional court with powers Japan’s Supreme Court exercises far more sparingly, on a remedy with a default outcome attached, and on a governing majority in a particular year. A reader who takes from this essay the instruction that Japan should simply wait for a court has read it backwards: the argument is that the Taiwanese court could act because of how the Taiwanese system is built, and that the same design is not lying around waiting to be used elsewhere.

It does not claim that the referendum was illegitimate. It was held under a law the legislature had just liberalised, it cleared a threshold set to be demanding, and more than seven million people voted in it. That it produced a result this essay reads as a cost does not make it a fraud, and treating an inconvenient majority as one is a habit this chamber has criticised elsewhere and will not adopt here.

It does not claim that the people who organised against the change were a foreign implant or a fringe. The coalition was religious in its organisation and a minority in its beliefs, and it nonetheless assembled a majority of those who voted, which means it was speaking to something in the country that was already there. Naming the churches explains the organisation. It does not explain the seven and a half million.

It does not claim a verdict in Japan. The Grand Bench has not ruled and a careful essay written in September 2026 cannot say what it will hold, how it will treat the split below it, or what the Diet will do afterwards. Nor does it claim that the partnership systems are worthless; they have altered the texture of ordinary life in ways the litigation has not, and they remain without force in national law. Both are true.

And it does not claim to have measured a society. The figures here are from named instruments with named limits, two of which disagree with each other. The account of what happened in Taiwanese classrooms after 2018 is read from the public record and from the accounts of people who work in them rather than counted. The essay is written by an outsider to both countries it compares, from the public record, in a language that is not the first language of either. This chamber’s essay on monogamy by choice argues that a default and a decision are different things even when they produce the same arrangement; the Library’s essay on imagining a life outside marriage argues that the shape of a life should be chosen rather than inherited. Taiwan adds the institutional version of both: that the form your relationship is permitted to take is decided by somebody, somewhere, on a date, and that it is worth knowing who and when.