Australia: A Vote on Someone Else’s Family, and What the Asking Cost

At half past eleven on the morning of 15 November 2017, the Australian Statistician announced that 7,817,247 people had answered Yes and 4,873,987 had answered No to the question of whether the law should be changed to allow same-sex couples to marry. Nearly eight in ten enrolled Australians had returned a form. Every state and territory had returned a majority Yes. So had 133 of the 150 federal electoral divisions. Three weeks later the Marriage Act was amended, and the thing was done.

That is the story as Australia tells it, and there is nothing dishonest in the telling. The margin was not narrow, the participation was not thin, and the result was not manufactured. People who had waited a long time got what they had waited for, and they celebrated in the street.

This essay is about the sentence the celebration leaves out. Before the country said yes, the country was asked — and a population being asked, for three months, in advertising and on television and across dinner tables, whether a particular group should be permitted to have families is not a neutral instrument that returns a number and leaves no residue. It returned a number and left a residue, and the residue was measured. That is the unusual thing about the Australian case: the cost of the procedure is not a matter of supposition. It is in the journals.

The chamber asks what becomes visible about lives here when another society is examined closely. Japan has never held a vote of this kind and is not about to; recognition has moved instead through ward offices and courtrooms. This essay argues that the different procedure carries a different bill, sent to different people, in different amounts. It does not argue that one bill is smaller.

The Marriage Act 1961 said nothing explicit about the sex of the parties until 2004, when Parliament inserted a definition confining marriage to a man and a woman and blocking recognition of same-sex marriages performed overseas. In 2013 the Australian Capital Territory legislated for same-sex marriage and the High Court struck the territory law down within days, holding that the federal Parliament had occupied the field. From then on the question was a federal one and only a federal one.

The government of the day had promised a national plebiscite. The Senate declined to authorise it, defeating the enabling bill on 7 November 2016 and a second attempt on 9 August 2017. On that same day the executive found another route. The Treasurer directed the Australian Statistician to collect what the governing statute calls statistical information: the views of enrolled Australians on whether the law should change. The 122 million dollars required was drawn not from an appropriation Parliament had voted for the purpose but from the Advance to the Finance Minister, a standing provision permitting expenditure the Minister considers urgent and unforeseen.

Two challenges reached the High Court at once. In Wilkie v Commonwealth and Australian Marriage Equality v Cormann the Court pronounced orders on 7 September 2017 and delivered unanimous reasons on 28 September, holding the appropriation authorised and what was being collected within the statutory concept of statistical information. The survey was lawful. Nobody put to the Court the separate question of whether it was wise, because that is not a question a court answers.

The design that resulted was peculiar. In a country where voting at elections is compulsory, this was voluntary. In a country with a professional electoral commission, it was run by the statistics agency. It bound nobody. Parliament had twice refused to authorise a national vote on the subject, and a national vote took place anyway, classified as a survey. Whatever one thinks of the outcome, the mechanism was a workaround, and the people whose lives were its subject had no say in choosing it.

The figures deserve to be stated exactly, because an essay that questions a procedure owes the result its full weight. Forms went out from 12 September 2017 to 16,006,180 eligible people. Responses closed at six in the evening on 7 November. 12,727,920 forms came back, a participation rate of 79.5 per cent. Yes took 61.6 per cent of responses, No 38.4 per cent. The exercise was budgeted at 122 million dollars and cost about 80.5 million.

Parliament then moved with a speed it had not managed in thirteen years of trying. The Senate passed the Marriage Amendment (Definition and Religious Freedoms) Bill on 29 November 2017 by forty-three votes to twelve, the House of Representatives on 7 December by one hundred and twenty-eight to four. It received assent on 8 December and commenced on 9 December. Marriage in Australian law became the union of two people to the exclusion of all others, voluntarily entered into for life. The same Act protected ministers of religion and defence force chaplains from any obligation to solemnise, created a new category of religious marriage celebrant able to decline, and exempted bodies established for religious purposes from providing facilities and services.

None of this reads as a technicality squeezed through on a bad turnout. A voluntary postal exercise drew participation many electoral systems would envy for a compulsory one, and older Australians participated at higher rates than younger ones, which had been the argument for expecting a different answer. If the question was whether the Australian public supported the change, the public answered clearly, and the essay does not dispute the answer.

What the essay disputes is the habit of letting the answer stand in for an account of the asking.

Ten days before forms went out, Parliament passed the Marriage Law Survey (Additional Safeguards) Act 2017. It commenced on 14 September 2017 and its protections ran until the result was declared on 15 November, at which point they lapsed.

What it did is the most eloquent document in the whole affair. It created temporary offences of vilifying, intimidating or threatening to cause harm to a person because of views they held or were believed to hold about the survey, or because of their religious conviction, sexual orientation, gender identity or intersex status. It prohibited bribery and deceptive conduct about the completion of forms, and required that communications carry authorisation particulars, so that anyone making a public claim could be identified.

Read the sequence again. A Parliament that had spent over a decade unable to agree about marriage agreed, within days and with little difficulty, that the campaign it was about to unleash would require a special criminal statute to be survivable — and that the statute could be switched off the moment the counting stopped. It is difficult to write a cleaner admission that the procedure was expected to produce conduct the ordinary law did not reach. The safeguards were not a concession wrung from advocates after the fact. They were built into the design, in advance, by the people who designed it.

They were also symmetrical, and this matters. The protections covered people of religious conviction as well as people of minority sexual orientation and gender identity, and there is a record of ugly conduct on both sides. An essay that counted only one direction of harm would be doing the thing it accuses others of doing.

The harm was not discovered afterwards. It was predicted, in writing, by the bodies whose professional business it is to predict such things, and the prediction was on the public record before a single form was printed.

The Australian Psychological Society opposed the use of a public vote, arguing that the empirical evidence pointed to damaging rhetoric in the resulting debate and to a risk of detrimental effects for lesbian, gay, bisexual, transgender, intersex and queer people. The Australian Medical Association issued a position statement on marriage equality in 2017 addressing the health consequences of exclusion from marriage and of the debate about it. Submissions to the parliamentary inquiries on the earlier plebiscite bill made the same case at length.

The evidence they drew on was mostly American. Through the 2000s a number of United States jurisdictions ran ballot measures on marriage, and researchers found elevated psychological distress among sexual minority residents of states running such campaigns. That literature has its own limits, and the professional bodies were extrapolating from one country to another. But extrapolating is what a professional body is for, and the extrapolation pointed in the right direction.

This sequence — warning, then survey, then measurement confirming the warning — is what makes the Australian case usable. It removes the defence that nobody could have known. Somebody did know, said so, was heard, and was overruled by a government that had run out of other ways to keep a promise.

Four peer-reviewed Australian studies measured the survey period, and they should be read with their weaknesses visible rather than summarised into a slogan.

Ecker, Riggle, Rostosky and Byrnes, in the Australian Journal of Psychology in 2019, recruited 5,742 LGBTIQ participants and 1,648 allies online during the survey period. Stress related to the debate predicted psychological distress in both groups, and continued to do so after accounting for prior life stress and pre-existing conditions. A companion qualitative analysis by the same team worked through 5,313 written responses. Participants described the period as pervasive and predominantly negative: prejudice encountered directly, fear, anger and depression, damage to relationships with family, colleagues and churches, and in some accounts thoughts of suicide.

Verrelli, White, Harvey and Pulciani, in Australian Psychologist in 2019, surveyed 1,305 lesbian, gay and bisexual Australians and found that more frequent exposure to negative media messages was associated with greater distress, the association significantly weaker among those who believed their own immediate circle supported them. Casey, Wootton and McAloon, in the American Journal of Orthopsychiatry in 2020, followed 2,220 LGBTQ participants to one week, three months and twelve months after the result. Distress and minority stress fell once the campaign ended, and greater exposure to it was associated with greater distress and perceived stigma, though not internalised stigma.

Now the limits. Every one of these samples was self-selected, recruited through social media and community networks, which means the people most affected by the campaign were also the most motivated to answer a questionnaire about it. Most designs are cross-sectional, establishing association and not direction, and none has a clean pre-campaign baseline drawn from the same individuals. No study can separate the survey from everything else that happened to those people in late 2017. Anyone citing these papers as proof that the survey caused a specified quantity of harm to a specified number of Australians is overreading them, and this essay does not.

What narrows the gap is a study from another country with a different method. Eisner, Fischer, Juster and Hässler, in the Proceedings of the National Academy of Sciences in 2024, tracked Swiss participants before, during and after the 2021 national vote on marriage equality, using preregistered analyses and hair samples. Hair cortisol and cortisone are not self-reports; a participant cannot decide to have more of them because the campaign upset her. Biological stress rose during the campaign among LGBTIQ people and those close to them, with the opposing campaign implicated and the supporting one buffering it, at an apparent cost to those doing the advocating. A different country, a binding referendum rather than a survey, and the same shape. That does not validate the Australian figures. It makes them harder to dismiss as an artefact of who volunteers.

Here is the thing this essay exists to say, and it will not be resolved into a verdict, because resolving it would falsify it. The result was a genuine and broad majority. It is also true that for three months an entire country conducted an argument, in paid advertising and on the evening news and at the front door, about whether a particular kind of family should be allowed to exist — and that the people who were its subject could not switch it off. They could decline to campaign. They could not decline to be campaigned about. A gay adolescent in a house where the form was filled in and marked No did not experience a national conversation. He experienced his own kitchen table, and no safeguards statute reaches a kitchen table.

Winning does not undo that. A majority verdict on your own legitimacy is still a verdict, and it leaves behind the fact that the question was live and answerable, and that 4,873,987 people — neighbours, colleagues, relatives, unidentifiable individually and permanently present in aggregate — answered it the other way. The couple who married in January 2018 know the national figure. They do not know which of the guests were in it.

This chamber has argued, in the essay on monogamy by choice, that an arrangement freely chosen and one passively inherited can look identical from outside and be entirely different from inside. The Australian case is the public version of that finding: a right conferred by a vote and the same right conferred by a statute are identical in the register and different in the body of the person holding them, because one was preceded by a period in which the answer was not yet known. And in the essay on when one person is not the plan, this chamber has held that recognition and legitimacy are separable goods. Australia separated them experimentally and re-fused them in a single month, and the separation is where the cost sat.

None of which establishes that the survey should not have happened. Without it, the Act very likely does not pass in December 2017, and every year of delay has its own cost, borne by the same people. The essay is not claiming the price was too high. It is claiming there was a price, that it was paid by the people the question was about rather than by those who designed it, and that it is left off the receipt almost every time the story is told.

The best argument against everything above is not made by opponents of marriage equality. It was made from inside, by Simon Copland in Australian Feminist Studies in 2018, and it deserves stating at full strength.

Copland argued that opposition to holding a public vote rested on a picture of LGBTIQ Australians as constitutively vulnerable and of the voting public as a dark and dangerous mass. Without denying the homophobic rhetoric that occurred, he argued that this picture was overstated, that campaigning against the vote reinforced the vulnerability it described and deepened a reliance on the state for protection, and that it ignored what a public argument can do — oblige families to declare themselves, put the question in rooms where it had never been raised, and reduce homophobia by the ordinary mechanism of exposure. On this reading the survey was a democratic event with a democratic yield, and the yield showed up in the result.

The research does not straightforwardly refute him. Casey and colleagues found distress declining after the campaign closed, not persisting. Verrelli and colleagues found that perceived support from one’s own circle measurably buffered the effect, which means the campaign also produced something protective. Ecker and colleagues recorded participants who described unexpected solidarity, and the Swiss study found the same buffering. The largest visible demonstration of support many LGBTIQ Australians had ever seen came out of the same three months that produced the harm, and it came because the question was asked out loud.

The essay takes the point and declines the conclusion. That an ordeal produces solidarity is not an argument for arranging ordeals, and the distribution matters: those best placed to gain from a public argument are the ones already surrounded by support, which is precisely the group Verrelli found was insulated. But Copland is right that treating a minority as too fragile for democratic argument is its own kind of diminishment, and a reader who takes from this essay that populations should be protected from hearing about each other has taken the wrong thing.

The essay’s own limits belong here. It reads a procedure from outside, in a country the writer does not live in. It relies on four studies with self-selected samples and one foreign study with a small biological subsample. It makes no claim about how many individuals were harmed. And it says nothing about the religious protections in the 2017 Act, a contested question a different essay would have to take on its own terms.

Japan has never put this to a vote and there is no mechanism by which it would, short of the one route this section ends on. Recognition has advanced along two tracks, neither of which involves the general public deciding anything.

The first is municipal. In 2015 Shibuya and Setagaya wards in Tokyo began issuing partnership certificates, and the practice spread. A joint survey by Shibuya Ward and the certified nonprofit Nijiiro Diversity found that as of 31 May 2025 there were 530 municipalities operating such systems, covering 92.5 per cent of the population, with 9,836 registrations recorded; the ward announced this would be its final survey, the system having become general. What these certificates cannot do is create a marriage. Inheritance, spousal taxation, joint parental authority and a foreign partner’s residence status are national matters, and no ward office can confer them.

The second is judicial. Six suits were filed from 2019. Five high courts — Sapporo in March 2024, Tokyo in October 2024, Fukuoka in December 2024, Nagoya and Osaka in March 2025 — held the absence of any provision for same-sex marriage to be unconstitutional, variously under the equality guarantee of Article 14 and the marriage and family provisions of Article 24. A differently constituted bench of the Tokyo High Court held on 28 November 2025 that the provisions were constitutional, its presiding judge adding that if the present situation continues a constitutional violation is inevitable and the Diet should deliberate the matter properly. On 25 March 2026 the Third Petty Bench of the Supreme Court referred six appeals to the Grand Bench. At the time of writing it has not ruled, and this chamber states no outcome.

Article 24 provides that marriage shall be based only on the mutual consent of both sexes. The Sapporo High Court held in March 2024 that the provision is not confined to its wording and protects marriage between people of the same sex to the same degree. On that reading nothing needs amending. On the opposing reading an amendment is required — and Article 96 requires that any amendment be submitted to the people at a referendum. That is where the two countries meet: the only route by which Japan would put this question to a national vote is the route that treats it as constitutional, and whether that route is necessary is the argument now before the Grand Bench. Meanwhile a nationally representative survey conducted in 2019 by stratified random sampling, with 2,632 valid responses from people aged twenty to seventy-nine, found 64.8 per cent agreeing that the law should recognise marriage between people of the same sex. Nobody was asked to decide. The deciding has been done by judges, and before them by a small number of plaintiffs who put their names and family histories into the public record in 2019 and are still waiting.

The comparative finding is about who is asked to decide, and what the asking itself does. Australia asked everyone at once, had its answer in three months, and the cost fell in a concentrated burst on the people the question was about. Japan has asked almost nobody, has taken over a decade, and is not finished; its cost is thin across time, uneven by postcode, since whether a relationship can be acknowledged at all still depends on which ward office one lives near, and extremely heavy on a handful of named individuals who have spent their thirties and forties litigating their own households. This chamber will not rank these. A vote is not more democratic than a ruling in any way that helps a woman whose mother stopped speaking to her when the form arrived, and a ruling is not gentler than a vote for a couple who cannot say when their case ends. The essay on the districts found that every country decides something and then arranges not to look at what it decided; every procedure sends a bill, and the recipient is rarely the one who chose the procedure.

This house should state its interest before anything else. It sells evenings organised around an arrangement that nobody certifies: no register, no ward office, no neighbour’s opinion, no form arriving in the post. It benefits, commercially and directly, from the existence of relationships and desires that have no public standing, because a woman whose attachment cannot be named at her own family table is a woman for whom a room that asks nothing has a price. That is the trade, and it should be said in the house’s own words rather than discovered by a reader.

What follows is a set of things the house may not claim. It may not claim that a private room is any kind of substitute for legal recognition. The difference between the two is inheritance, hospital corridors, immigration and children, and no evening addresses any of them. It may not claim that discretion is a virtue when discretion is a product. It may not claim that its customers are in any particular situation, because it does not know and has no business guessing.

And it may not claim the most flattering thing: that an evening in which nobody asks you to justify yourself is the same as the right not to be asked. It is not. It is a few hours of the condition, bought, bounded and ending at a known time. That is real and worth what it costs, but the difference between renting the condition and holding it is the entire subject of this essay, and a house that blurred it would be selling the wrong story about its own product.

It does not claim that the postal survey should not have been held, or that it should have been. The counterfactual is genuinely uncertain and the essay refuses both halves of it.

It does not claim a prevalence. The studies cited measure associations in volunteer samples, and this essay makes no assertion about how many LGBTIQ Australians were harmed, how severely, or for how long. The Swiss biomarker study corroborates a direction, not a quantity, and it describes a different country.

It does not claim that people who voted Yes bear responsibility for what the campaign cost, or that people who voted No were all engaged in the conduct the safeguards statute was written to reach. It says nothing about any individual voter, campaigner, litigant or public figure in either country, and nothing about the religious protections in the 2017 Act beyond noting that they exist and are contested.

It states no outcome for the six appeals before the Grand Bench of Japan’s Supreme Court. It does not predict what the Diet will do. Where a matter is still moving, it is described as moving.

It contains no diagnosis of any reader, no generalisation about the character of either nation, and no endorsement of any organisation, party, campaign or plaintiff. It is a comparison of two procedures and of who pays for each, offered as a mirror rather than as a recommendation, and it declines to say which mirror is kinder.